We present a preview version of DeepSeek-V4 series, including two strong Mixture-of-Experts (MoE) language models -- DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) -- both supporting a context length of one million tokens. DeepSeek-V4 series incorporate several key upgrades in architecture and optimization: (1) a hybrid attention architecture that combines Compressed Sparse Attention (CSA) and Heavily Compressed Attention (HCA) to improve long-context efficiency; (2) Manifold-Constrained Hyper-Connections (mHC) that enhance conventional residual connections; (3) and the Muon optimizer for faster convergence and greater training stability. We pre-train both models on more than 32T diverse and high-quality tokens, followed by a comprehensive post-training pipeline that unlocks and further enhances their capabilities. DeepSeek-V4-Pro-Max, the maximum reasoning effort mode of DeepSeek-V4-Pro, redefines the state-of-the-art for open models, outperforming its predecessors in core tasks. Meanwhile, DeepSeek-V4 series are highly efficient in long-context scenarios. In the one-million-token context setting, DeepSeek-V4-Pro requires only 27% of single-token inference FLOPs and 10% of KV cache compared with DeepSeek-V3.2. This enables us to routinely support one-million-token contexts, thereby making long-horizon tasks and further test-time scaling more feasible. The model checkpoints are available at https://huggingface.co/collections/deepseek-ai/deepseek-v4.

While In-Context Learning (ICL) is extensively studied in Autoregressive (AR) LLMs, its mechanism within Diffusion Large Language Models (dLLMs) remains largely unexplored. Unlike AR models restricted by unidirectional causal masking, dLLMs intrinsically utilize bidirectional attention, offering extensive spatial flexibility for query placement. Unfortunately, current practices conventionally inherit AR-style trailing-query templates, often overlooking the structural paradigm shift. This paper presents a comprehensive analysis unveiling that query position is actually a first-order variable in dLLMs. Through empirical decoupling, we demonstrate that positional variance impacts generation quality on par with example semantic quality. Internally, this positional sensitivity stems from a spatial ``Recency Effect'' in attention flow and task-dependent shifts in decoding trajectories. To mitigate this instability without ground-truth labels, we reveal that traditional single-step confidence ($C_{decoded}$) fails in dLLMs. Instead, we propose Average Confidence ($\overline{C}$), a novel metric tracking the iterative decoding process. By establishing the foundational spatial ICL baselines, we introduce Auto-ICL, a training-free adaptive routing strategy that dynamically optimizes query placement, robustly approaching oracle performance across heterogeneous reasoning and perception tasks.

Large language models (LLMs) excel at multi-step reasoning but incur substantial inference cost. We introduce Causal Attribution Pruning (CAP), a training-free method that identifies critical attention heads by measuring their causal impact on reasoning tasks and uses these head-level scores to guide fine-grained weight pruning. For each attention head, CAP estimates the expected performance degradation when the head is masked during forward passes on a small calibration set of reasoning problems. These causal scores are then converted into weight-level importance values for the corresponding projection matrices. Unlike magnitude-only or activation-based criteria, CAP's interventional measurement directly captures each head's functional contribution, yielding relative accuracy gains of up to 61% over Wanda on ARC-Challenge at 20% sparsity. We evaluate CAP on GSM8K, StrategyQA, and ARC-Challenge using Llama-3-8B-Instruct and Mistral-7B-Instruct at 10%, 20%, and 50% sparsity. At moderate sparsity (10-20%), CAP improves over Wanda in most model-benchmark configurations. with especially large gains on ARC-Challenge for Llama-3. Our results suggest that attention-head-level causal attribution can better preserve reasoning performance on downstream benchmarks than correlational pruning criteria at equivalent sparsity, while remaining limited by coarse MLP attribution at 50% sparsity.

In-Context Learning (ICL) allows LLMs to adapt to new tasks from a few demonstrations, but its reliability remains a concern: predictions are highly sensitive to both prompt design and the model's ability to understand the context, obscuring whether failures arise from data properties or model limitations. Uncertainty decomposition-separating aleatoric from epistemic sources-is particularly crucial in this setting, yet existing methods, designed for standard generation tasks, fail to capture the unique dynamics of ICL. To address this, we introduce a concept of self-function vectors, built upon Bayesian views and the mechanistic interpretability of ICL. These vectors leverage internal model representations to model the latent concept learned during in-context prompting, thereby enabling a direct estimation of aleatoric uncertainty within a Bayesian framework and circumventing the reliance on brittle input or decoding manipulations. Given the lack of established benchmarks and suitable evaluation protocols, we also propose the first and rigorous evaluation protocol, in which data is manipulated in controlled ways so as to quantify aleatoric uncertainty precisely and separately from epistemic uncertainty. With this new evaluation framework, initially grounded in synthetic tasks for conceptual development and subsequently extended to real-world datasets, we show that our proposed methodology can measure uncertainty of LLM predictions made under ICL more reliably than existing alternative methods. Moreover, we show it can be used as a practical tool for trustworthy-related applications, such as hallucination detection. Our findings pave a new direction for connecting the quantitative view of uncertainty with the mechanistic understanding of model behavior.

Test-time scaling (TTS) has emerged as a powerful paradigm for improving the reasoning performance of large language models (LLMs) by investing additional compute at inference time. A central component of TTS is the \emph{verifier}, which selects or scores candidate solutions to guide the search process. While prior work has explored the benefit of verification, a fundamental question remains underexplored: \emph{what is the optimal granularity of verification under a given compute budget?} Coarse-grained outcome reward models (ORMs) and fine-grained process reward models (PRMs) represent two extremes, yet neither alone achieves compute-optimality across all regimes. In this paper, we establish a unified theoretical framework, called \textbf{GRACE} (\underline{G}ranularity-\underline{R}egulated \underline{A}daptive \underline{C}omputational \underline{E}fficiency), that characterizes the optimal verification granularity as an explicit function of problem difficulty, verifier accuracy, and compute budget. We prove that there exists a phase transition: fine-grained verification dominates when either the compute budget is large or the problem is hard, whereas coarse-grained verification is preferred in the low-budget, easy-problem regime. Our theory unifies Best-of-$N$, beam search, and step-level MCTS within a single Pareto-optimality framework, and motivates an adaptive granularity strategy that provably achieves the compute-performance Pareto frontier. Empirical results on MATH-500, GSM8K, and AIME benchmarks corroborate all four theoretical claims, with our adaptive strategy outperforming fixed-granularity baselines by up to 3.1\% accuracy at matched compute.

We use training-data attribution as an interpretable tool for capability discovery, mapping which regions of the pretraining corpus support social-reasoning versus STEM-reasoning in OLMo3-7B. Training-data attribution measures how strongly each training document influences a model's predictions on a benchmark, but document-level scores are too noisy to identify which corpus regions support which capabilities, and prior work has emphasized factual knowledge rather than reasoning. We compute gradient-based attribution (TrackStar via Bergson) over a working set drawn from the de-duplicated Dolma3 mix, aggregate influence across WebOrganizer's 24-format x 24-topic taxonomy (576 bins), and contrast benchmark pairs in a 2x2 design that varies domain (social vs. STEM) and capability type (reasoning vs. knowledge): SocialIQA and MMLU Social Sciences against ARC-Challenge and MMLU STEM. Social and STEM reasoning draw on qualitatively distinct corpus regions, and the contrast is sharper at the reasoning level than at the knowledge level. Targeted machine unlearning provides partial causal validation: forgetting high-attribution topic bins (e.g., Literature for SocialIQA) degrades the aligned benchmark more than within-bin random baselines, and we open-source all code, sampling manifests, the bin-level influence matrix, and unlearning checkpoints.

Multilingual Large Language Models (MLLMs) are increasingly expected to handle Code-Switched (CS) inputs, yet mixing languages frequently degrades performance relative to source- or target-language monolingual counterparts. To understand this degradation, we use grammar-forced CS as a controlled diagnostic setting for locating CS representations relative to their source and target counterparts. We introduce Anchor Bias, a geometric measure that quantifies language anchoring, whether a CS hidden state aligns closer to its source or target language counterpart. Across diverse MLLMs, Anchor Bias reveals a consistent grammar-frame effect: source-framed CS stays source-anchored, whereas target-framed CS shifts target-ward and shows larger Question Answering (QA) degradation. Motivated by this representational pattern, we propose CANVAS (Contextual Anchor-based Neural Vector Alignment Steering), an inference-time intervention that extracts a source-side canvas from the input and softly steers target-language hidden states toward the source anchor during prefill. CANVAS consistently recovers QA F1 across MLLMs and CS conditions, showing that internal anchoring signals provide an actionable target for mitigating CS inference failures.

Aligning language models with human preferences often requires optimising multiple behavioural objectives. A practical approach is to apply these objectives sequentially using preference optimisation methods such as Direct Preference Optimisation (DPO), but it remains unclear whether later training uniformly degrades preferences learned earlier or whether the effect depends on the relationship between objectives. We study sequential DPO across four preference settings covering distributional conflict, multi-attribute interaction, strong safety signal, and compatible response-quality objectives. Using Llama-3.1-8B-Instruct with LoRA adapters, we evaluate all objectives after every stage with a fixed base-model reference. We find that sequential DPO does not produce a single forgetting pattern; preference change ranges from partial degradation to stability, pair-level redistribution, or positive transfer depending on objective relationship, signal strength, and training order. Pair-level analysis using length-normalised policy margins shows that aggregate metrics can mask heterogeneous changes across preference pairs, whereas quartile decomposition reveals that high-confidence pairs can either degrade or improve depending on the setting. Mechanistic diagnostics show that Stage~2 gradients and adapter updates are near-orthogonal to the previous objective across all settings, providing little evidence that direct gradient opposition is the primary driver. These findings suggest that future sequential alignment pipelines should account for objective compatibility and signal strength, rather than assuming that later objectives affect earlier preferences uniformly.

Pre-trained language models such as BERT achieve strong text classification performance but lack transparency, limiting their use in high-stakes settings. The Tsetlin Machine (TM) offers fully interpretable, clause-based reasoning but captures little semantic information, and prior attempts to bridge the two rely on static word embeddings that miss contextual meaning. We propose a semantic pre-training framework that transfers knowledge from a pre-trained language model into a TM without using embeddings. Text samples are grouped into semantically coherent clusters with K-means or Top2Vec, and the resulting cluster-sample pairs pre-train a non-negated TM with enhanced Type I feedback. The TM thereby learns interpretable semantic keywords that are fine-tuned on downstream tasks. Across five datasets, our method substantially outperforms vanilla and embedding-based TMs and reaches performance competitive with BERT while remaining interpretable.

Aligned language models gate behaviors such as refusal and language routing through sparse feed forward neurons, yet no theory predicts when a single neuron intervention controls a behavior coherently rather than collapsing the output. We develop a budget normalized control window framework for single neuron steering. A dose along one write direction reduces to one control coordinate: the alignment between the residual stream and the write, driven along a universal saturation curve in units of a coherence budget set by the residual norm divided by the write norm. Coherent control exists when a behavior trigger lies below the collapse ceiling. The same coordinate governs benign mode switches and refusal; the ceiling follows from weights and one generic forward pass, while triggers are measured at rollout. On fifteen held out neurons, the predicted ceiling has mean absolute error 0.14, about 0.07 in bulk layers, and the committed open or closed verdict holds on eleven against a ten of fifteen majority baseline. Closed cases expose three failure modes rather than violations: collapse before trigger, too little depth to propagate, or a normalization that caps how far one neuron can push. The law explains why local gradient attribution anti predicts control: true controllers write off the readout axis and carry a near zero first order gradient. A forward only contrastive screen made precise by the window recovers controllers that attribution misses. On refusal, the hardest case, intervention success is typed, not scalar: coherent bypass and strict actionable reach separate, so a neuron can flip refusal in fluent, on task text with no actionable content, and genuine actionable reach appears only for three of six audited Llama pivots and only at later rollout horizons. Single neuron steering is therefore a budgeted, typed audit of controllability rather than a fixed dose anecdote.

Large language models (LLMs) are commonly prompted and interfaced with human-readable natural language, even when the intended reader is another model. This paper investigates whether semantic information can be encoded in compact, non-standard textual forms that sacrifice human readability while remaining recoverable by LLMs. We refer to this class of model-centric textual representations as BabelTele, approached here not as a fixed protocol but as an empirical probe into LLMs' capacity to generate and interpret such representations. Through readability diagnostics, model likelihood measures, human questionnaires, and downstream task evaluations, we find that BabelTele can substantially depart from ordinary natural language while preserving core semantics for instruction-tuned LLMs. As a task-agnostic representational paradigm, BabelTele demonstrates high information density, maintaining 99.5% semantic fidelity even when the text volume is condensed to 27.9% of its original length. We further evaluate its semantic robustness in cross-model transfer, agent memory, and multi-agent communication. Results suggest that BabelTele can reduce context overhead while generally maintaining reliable downstream performance, although its effectiveness depends on the compressor-reader pair and task setting. These findings indicate that human readability, natural-language typicality, and model-side semantic recoverability can be partially decoupled, opening a path toward model-native representations in future exploration of LLM systems.

Activation steering controls model behavior by modifying intermediate hidden states at inference time without retraining. Existing methods handle only single-direction injection; when multiple semantic directions are superposed without constraints, the model collapses. We show that this collapse decomposes into two independently acting sources: distributional deviation, where additive perturbations accumulate in norm across layers and drive activations outside the training distribution, and directional interference, where non-orthogonal semantic vectors mutually dampen when superposed. These two sources define the design constraints that any training-free multi-directional intervention must address. As one instantiation of these principles, we propose GEMS, a training-free method that maps each source to a corresponding geometric constraint: norm-preserving weighted superposition and targeted attention-pathway injection for distributional deviation, and real-time orthogonalization for directional interference. On GSM8K, injecting three concurrent non-mathematical directions preserves accuracy at 98% (baseline 92%), while unconstrained addition collapses to 4%; on Wikitext-2, the same injection incurs only 2.2% PPL increase. Component ablation isolates the causal role of each constraint, and layer-level probes confirm that orthogonalized signals survive the FFN pathway and reach the output distribution with semantic specificity. Qualitative steering effects transfer across architectures from 3B to 31B.

Persian pretrained language models (PLMs) are still limited by the scarcity of large-scale, high-quality pretraining corpora and by insufficient evaluation beyond standard classification and NER tasks. We present IHUBERT, a monolingual Persian PLM trained from scratch with the RoBERTa-base encoder (125M parameters) on a 45 GB curated subset of the Sepahr-Danesh collection (about 7-8B tokens). To improve corpus quality and reduce redundancy, we employ a multi-stage preprocessing pipeline that includes normalization, exact and near-duplicate removal, anonymization, and vector-database-based semantic deduplication for distribution balancing control across domains and registers. We additionally train a 139k-vocabulary BPE tokenizer on the full pretraining corpus to better capture Persian morphology and orthographic variation. IHUBERT is evaluated on seven Persian NLU benchmarks covering NER, sentiment analysis, topic classification, NLI, extractive question answering, and relation extraction, using task-standard metrics (entity-level F1, Macro-F1, EM/F1). IHUBERT achieves its strongest gains on extractive QA, ranking first on both PQuAD (F1 88.3542) and ParsiNLU-RC (F1 49.0987), and attains the best result on FarsTail (Macro-F1 0.8350). On NER and topic classification, it remains competitive (e.g., 0.8308 F1 on ParsTwiNER; 0.7953 Macro-F1 on DigiMag), while relation extraction remains the main remaining gap (0.6684 Macro-F1 on PERLEX). A controlled tokenizer ablation on the IHUBERT pretraining corpus shows that BPE yields slightly lower subword fragmentation than WordPiece at matched vocabulary size, supporting our tokenization design. Overall, IHUBERT advances Persian language modeling through semantically curated large-scale pretraining and broad evaluation across both classification and comprehension-oriented tasks.

The quadratic complexity of attention poses a critical bottleneck for long-context processing, spurring interest in hybrid attention designs. Most open-source hybrid models adopt a layer-wise strategy. Yet, prior work has noted the inherent difficulty of integrating Linear Attention (LA) with Full Attention (FA), suggesting that the design space of attention hybridization remains underexplored. To probe this space, we conduct interpretability analysis and observe that layers exhibit block-wise functional similarity, while individual heads within the same layer display distinct functional specialization despite sharing input features. This head-level heterogeneity suggests that the head dimension provides a natural and principled granularity for fusing heterogeneous attention signals. Building on this insight, we introduce HydraHead, a novel architecture that hybridizes FA and LA along the head axis. HydraHead features two key innovations: (1) an interpretability-driven selection strategy that identifies retrieval-critical heads and preserves FA only for them, and (2) a scale-normalized fusion module that reconciles the distributional gap between FA and LA head outputs. By leveraging a three-stage transfer pipeline with parameter reuse and distillation, we achieve high-performance hybrid models with minimal training overhead. Under a unified training setup, HydraHead outperforms other hybrid designs in long-context tasks while maintaining strong general reasoning. With interpretability-driven head selection, it matches a 3:1 layer-wise hybrid's long-context performance at a 7:1 LA-to-FA ratio. Crucially, trained on only 15B tokens, HydraHead achieves over 69% improvement over the baseline at 512K context length, approaching Qwen3.5, a leading model of comparable size with a native context length of 256K. This highlights the significant scaling potential of head-level hybridization.

To align a Large Language Model (LLM), most existing methods collect explicit human feedback and train a reward model to predict the human preference based on the response text. These existing methods have two key limitations. First, the users rarely provide explicit feedback for LLM responses, which makes the high-quality preference annotation expensive to collect. Second, the methods do not leverage implicit human feedback, which has proven vital to the economic moats of Internet giants. To quantify the value of implicit feedback, we build a new dataset called IFLLM, which collects 1336 multi-turn questions from the 59 Mechanical Turk workers, their mouse trajectories, and eye gazing points to the LLMs' responses from their webcams. IFLLM shows that the users have very diverse types of gazing behavior and mouse trajectories. Our reward model based on the implicit user feedback boosts the accuracy of the text-based reward model from 55% to 64% and nearly triples the relative response quality improvements after applying the DPO to eight LLMs, demonstrating the value of implicit feedback in the wild. Our data collection website, dataset, and codes can be found at https://github.com/themehulpatwari/llm-implicit-feedback/.

We study cross-lingual transfer by fine-tuning seven large language models (4B--671B parameters) on Arabic and evaluating zero-shot reading comprehension on Semitic languages and non-Semitic controls. Across dense and Mixture-of-Experts architectures, we find no evidence of Semitic-specific transfer: models with weak baselines improve dramatically across all languages, while strong-baseline models show only marginal gains regardless of language family. A chain-of-thought ablation reinforces this finding -- the same models that benefit most from fine-tuning benefit equally from inference-time reasoning, suggesting both mechanisms address task-format alignment rather than cross-lingual knowledge transfer.

AI and large language models (LLMs) have emerged as promising tools to address global mental health challenges. Despite the global nature of these challenges, there remains a critical shortage of high-quality datasets for training and evaluating such systems. To mitigate this gap, researchers increasingly generate synthetic clinical personas to simulate user data and test digital mental health support systems. However, most validated personas rely on English-centric contexts. This paper investigates whether similar persona-based methods can be used to generate multilingual mental health datasets. We modified nationality and language parameters in personas to generate clinical dialogues in Mandarin, Bengali, and Hindi. We then examined how different LLMs perform when evaluating the depression severity of these generated multilingual datasets against the baseline in English. Our findings indicate that just adding nationality and language parameters in personas might not be adequate, as it can introduce clinical inconsistency across languages. LLM judge models often exhibit inaccuracies in assessing depression severity in non-English texts, with performance varying across different models. This exposes the systemic limitations of applying English-centric personas to multilingual contexts. Ultimately, our work highlights the urgent need for culturally responsive data generation to ensure equitable mental health systems globally.

The rapid increase in scientific publications leads to the fact that manual study screening in systematic literature reviews (SLRs) is increasingly resource consuming, inefficient, and inconsistent. Classifying studies that clearly report health-related quality-of-life results, such as EQ-5D data, requires a high level of clinical interpretation and poses challenges for human reviewers. This study investigates the use of Google's Gemini and Gemma large language models (LLMs) in automating EQ-5D detection in the PubMed biomedical database based only on published abstracts. A multi-phase framework is proposed that integrates few-shot prompting, weight ensembling aggregation, and a soft stacking meta-classifier. Nine LLMs are evaluated on a dataset of PubMed studies manually labeled by two experts regarding EQ-5D reporting. The weighted ensemble of gemini-2.5-pro, gemma-3-12b, and gemma-3-27b obtained a 0.74 weighted F1-score and 0.74 accuracy, exceeding individually attained results. The ensembling of top-performing models improved the balance between precision and recall compared to individual models, while the soft stacking approach provided greater reliability and interpretability. Feature analysis shows that the probability results from the models are important in guiding the final predictions. The findings suggest that an ensemble-based LLM setup is a reliable and scalable approach for automating screening in biomedical research.

Knowledge graph (KG) reasoning infers new knowledge from existing facts and is widely applied in question answering, recommendation, and decision support. With the rapid development of large language models (LLMs), LLM-based KG reasoning frameworks have become increasingly popular by leveraging retrieved KG information. However, hallucinations in LLMs remain a critical issue. Even when relevant KG knowledge is incorporated, models may still generate incorrect outputs, leading to misinformation and unreliable decisions. Existing hallucination detection methods either focus on LLM internal states or verify consistency with retrieved contexts, but both overlook the structural information in KGs, resulting in suboptimal performance. To address this gap, we propose LUCID, the first halLUcination deteCtIon method for LLM-based knowleDge graph reasoning frameworks. LUCID jointly leverages LLM attention scores, KG semantics, and structural information. Specifically, it extracts node and edge features from attention scores and semantic similarities, and integrates them with KG structure using a graph neural network. We also construct manually annotated benchmark datasets for evaluation. Experiments on nine datasets show that LUCID achieves state of the art performance compared to 15 baselines.

Retrieval-Augmented Generation (RAG) improves factual grounding, but it also lengthens prompts and raises prefill cost. Prefix caching in serving engines such as vLLM reduces this cost only when requests share the same token prefix. In grounded generation, however, adjacent queries may retrieve overlapping evidence in different orders, so set overlap does not become reusable prefix overlap. We present CacheWeaver, a lightweight prompt-layer method for cache-aware evidence ordering. The method keeps a prefix tree over recently served evidence sequences and uses a greedy walk to place the most reusable prefix first, while leaving the serving engine and retrieved evidence set unchanged. Across three vLLM configurations, the method lowers median time-to-first-token (TTFT) by about 20-33 percent relative to retrieval-order prefix caching, without hurting answer quality in our QA tests. The greedy policy reaches 97.5 percent of the median TTFT gain from oracle ordering, indicating that most reusable prefix locality can be recovered by a simple scheduling layer between retrieval and inference.

Researchers are interested in learning about Mars so that it may eventually become habitable for humans. To achieve this, there is a need for comprehensive knowledge of the planet's atmosphere, hydrology, surface chemistry, radiation environment, and spatial features through the scientific literature. These contain valuable information and meaningful quantitative constraints that can be used in other models and studies, such as habitability assessment and future terraforming studies. We present TerraMARS, an end-to-end information extraction pipeline that combines a domain-adapted Small Language Model to answer Mars terraforming-related questions and convert unstructured Mars science text into machine-readable structured outputs in JavaScript Object Notation (JSON) format. A corpus of open-access papers is collected and processed using a multistage retrieval and chunking framework. Google Gemma 3 1B was adapted to the domain using Quantized Low-Rank Adaptation (QLoRA) fine-tuning on Mars-specific question-answering and information extraction datasets. The resulting pipeline generates both types of output and provides a foundation for integrating knowledge from scientific literature into downstream applications like digital twins and habitability modeling for Mars. The output from this pipeline looks promising, but further improvements are needed to increase extraction accuracy and factual consistency.

Court proceedings contain valuable evidence about human smuggling networks, but this information is often buried within unstructured, jargon-heavy legal documents. While large language models (LLMs) can support knowledge graph construction through automated information extraction, existing approaches rely on general-purpose models that are not tailored to the entity and relationship definitions required in this domain. We introduce FineREX, a streamlined knowledge graph construction pipeline built around a fine-tuned LLM for named entity recognition and relationship extraction (NER-RE). Using a manually annotated dataset of $512$ text chunks, FineREX achieves absolute improvements of 15.50% and 31.46% in entity and relationship F1-score, respectively, compared to a larger general-purpose baseline. These gains translate into higher-quality knowledge graphs, reducing legal noise by nearly half and lowering node duplication on long documents from 17.78% to 11.17%. By eliminating document rewriting and redundant extraction stages, FineREX also reduces end-to-end processing time by 50.0%. Our results demonstrate that domain-specific fine-tuning can substantially outperform larger general-purpose models while improving both the quality and efficiency of knowledge graph construction for illicit network analysis.

Information extraction from pathology reports is essential for cancer staging, tumor registry population. Yet key data remains embedded in narrative reports, making manual extraction labor-intensive and error-prone. Traditional supervised Natural Language Processing pipelines address this through fully supervised Named Entity Recognition and Relation Extraction, but require expensive manual annotation and suffer cascading failures when upstream entities are missed. In this study, we developed a zero-shot, agentic workflow, and evaluated five open-source generative Large Language Models (LLMs) to populate 13 College of American Pathologists synoptic fields from lung resection pathology reports. We compared them against a state-of-the-art supervised GatorTron NER-RE baseline using a novel, registry-aligned evaluation framework. The baseline achieved Micro-F1of 0.960, while the best zero-shot model (GPT-OSS-20B) achieved Micro-F1 of 0.893 (recall: 0.949), accurately extracting complex relations like Pathologic Stage without task-specific training. These results suggest that open-source, zero-shot agentic LLMs are a low-cost solution for extracting lung pathology information.

From financial filings to clinical records, legacy industries rely heavily on long, unstructured documents to store high-value information. Reliably extracting this information into structured, machine-readable representations is a key prerequisite to making the contents accessible to automated systems. JSON is a natural target for such structured extraction, yet constructing reliable and scalable text-to-JSON training data remains challenging. To address this gap, we propose STAGE (Spreadsheet-grounded Text-to-JSON Artifact GEneration), a source-grounded data generation pipeline that constructs reports and JSON schema by using LLMs for scalable synthesis while validating ground-truth values against the underlying spreadsheet. Evaluations on STAGE-Eval, our source-grounded benchmark with an 851-example test set, show that STAGE produces stronger training data than existing approaches. This improves Qwen3-4B exact match from 31.37% to 74.27% and value accuracy from 45.46% to 90.69%.

Streaming Retrieval-Augmented Generation (Streaming RAG) reduces user-perceived latency by issuing tool queries in parallel with ongoing user input, before the utterance is complete. Reported gains are aggregate, yet the mechanism's benefit is fundamentally query-intrinsic: speculation can only help when the correct tool query becomes determinable before the user stops speaking or typing. We isolate and measure this property -- tool-intent stabilization, the point in the input stream at which a speculative query's retrieval converges to the answer-bearing result. On the CRAG benchmark (1371 validation questions) we (i) measure the distribution of stabilization, (ii) derive a model-agnostic bound H on the portion of tool latency that can be hidden behind the user's remaining input, as a function of tool latency L and input cadence δ, (iii) validate against a working streaming pipeline that realized savings meet or exceed this bound, and (iv) identify which query properties predict early versus late stabilization. The study requires no model training and runs on commodity CPU hardware. We find that at a realistic operating point (L=600ms, δ=3w/s, θ=0.8), 73.9% of queries across the full benchmark admit substantial latency hiding -- a blended figure that mixes sufficiency stabilization on the 21.3% of questions where gold evidence is verbatim-present and BM25-retrievable (95.2% streamable on this favorable slice) with a grounding-free top-1-settling fallback on the remainder. On the favorable slice, ϕ_suf is bracketed to [0.26, 0.281] by exact and relaxed grounding -- both early. Question type produces a significant but coarse early/late split (Kruskal-Wallis p=0.017, epsilon^2=0.04), directly informing when a learned speculative trigger is worth its cost.

When multi-agent LLM systems produce bad answers, not all failures are equal: some answers are grounded in the right material but incomplete, while others are simply ungrounded and should be stopped. Current retry strategies treat both cases identically (try again and hope for the best), leaving human supervisors unable to tell whether a retry was warranted or whether the system should have halted instead. We introduce the Argent Signaling Protocol (ASP), a compact machine-readable header that accompanies every AI-generated response with structured quality signals: certainty (@C), grounding (@G), stochasticity (@S), and an assumption index that classifies the evidentiary basis of each claim. These signals enable a controller to distinguish repairable failures from containment failures and route each case differently. We evaluate ASP in two modes. In standalone mode, a 27-question document-grounded QA benchmark over the Array BioPharma/Ono license agreement compares baseline prompts against ASP-instrumented controller actions across three local GGUF models. On Qwen~(0.8B), ASP improves pass rate from 11.1% to 33.3% and mean term coverage from 36.7% to 65.4%; on Dobby~(8B), ASP produces 4 fail-to-pass recoveries, raising pass rate from 33.3% to 44.4%; on SmolLM3~(3B), ASP alternates between repair and containment per question. Aggregate improvement is meaningful (12/81 to 21/81 passes). In multi-agent mode, an ASP sidecar sits between a retrieval agent and a downstream decision agent; the sidecar blocks 100% of ungrounded upstream outputs from reaching the downstream agent (24/27 blocked, 0 ungrounded propagations).

On-policy distillation (OPD) improves student models by training them on trajectories induced by their own policy, making it a promising approach for mitigating exposure bias in agent training. However, most OPD studies focus on single-turn settings, while realistic LLM agents interact with environments over multiple turns. In this regime, early errors can alter future observations and compound across the trajectory, and standard dense token-level OPD becomes brittle, as it may over-penalize semantically valid alternatives, reinforce local degeneracies such as repeated actions, and propagate unreliable teacher supervision on off-distribution histories. We propose SAGE-OPD, a verifier-free selective intervention framework specifically designed for multi-turn OPD. Instead of applying teacher supervision uniformly across all turns, SAGE-OPD first observes environment feedback and uses teacher judgment to decide whether each student response should be skipped or intervened on. To further address compounding errors, SAGE-OPD weights token-level distillation by teacher confidence, reducing the influence of uncertain teacher distributions on corrupted or ambiguous histories. Finally, SAGE-OPD applies loss normalization to preserve the overall loss scale of standard OPD while retaining selective turn-level weighting. Experiments on agent tasks show that SAGE-OPD consistently improves over baselines, achieving up to a 13.3% relative improvement in ALFWorld unseen success rate over standard OPD. Ablation studies further demonstrate that turn-level intervention, teacher confidence weighting, and loss normalization provide complementary benefits. Our results suggest that effective multi-turn OPD should remain on-policy, but teacher supervision should be selectively allocated to turns where intervention is necessary and reliable.

Large language models (LLMs) demonstrate strong reasoning and generation abilities, but their fixed context windows limit long-term information accumulation and reuse across multi-session interactions. Existing memory-augmented systems often construct memory in a coarse and unstable manner, relying on inefficient memory representations or unstable unconstrained updates. To address these challenges, we propose AtomMem, a long-term memory system designed for value-dense storage and stable memory evolution. AtomMem introduces a Fact Executor, which selectively extracts high value atomic facts from long form interactions to serve as highly efficient memory representations. Subsequently, AtomMem organizes these facts into hierarchical event structures and temporal profiles, capturing coherent episodic contexts and tracking dynamically evolving user attributes over time. During retrieval, the system activates an associative memory graph to connect fragmented memories. Experiments on the LoCoMo benchmark confirm that AtomMem achieves state-of-the-art performance across various reasoning tasks, offering a scalable and economically viable solution for deploying intelligent personalized agents.

Real-world computer-use tasks often span multiple applications and devices, requiring agents to coordinate heterogeneous environments under dynamic runtime failures. Existing multi-device agent systems support task decomposition and cross-device assignment, but recovery remains largely coarse-grained: when execution fails, they typically retry the same strategy, reassign the subtask, or revise the global plan, without systematically modeling the device-local strategy space. This limits their ability to distinguish failures that can be repaired within the current device from those that require cross-device replanning. We propose \textbf{H-RePlan}, a hierarchical replanning framework for multi-device agents with unified API--CLI--GUI execution. H-RePlan equips each device with interchangeable execution strategies and separates device-local strategy recovery from orchestrator-level global replanning through a compact cross-layer failure abstraction. To evaluate this capability, we introduce \textbf{HeraBench}, a fault-injected benchmark that constructs cross-device workflows over Linux and Android devices and injects strategy- and device-level failures. Experiments show that H-RePlan substantially outperforms single-strategy and coarse-grained multi-device baselines, achieving higher completion, instruction adherence, and perfect-pass rates while reducing the token cost required for reliable end-to-end success. These results demonstrate that scope-aware hierarchical recovery is essential for robust multi-device agent execution.

In this technical report, we focus on solving the challenge of Vietnamese multi-document abstractive summarization, introduced in the International Workshop on Vietnamese Language and Speech Processing (VLSP) 2022. We choose to follow the popular hierarchical approach, i.e. condensing each document followed by aggregation and summarization. We propose a novel yet simple strategy to shorten documents that is driven by the golden summary, thus ensuring high correlation between stages of the hierarchical approach. Our method achieves a ROUGE2-F1 score of 0.2468 on the VLSP's public test set, and can produce fluent and concise summaries. Additionally, we utilize external sources for extra data, which greatly enhances the quantity of data for Vietnamese multi-document summarization. The additional data is made available for the community.

Effective Automated Essay Scoring (AES) are expected to support both reliable assessment and actionable instructional feedback. However, existing approaches often treat scoring and feedback as separate components: neural scoring models provide limited interpretability, while Large Language Model (LLM)-based feedback is typically insensitive to learners proficiency levels. To address this fragmentation, this work proposes PsyScore, a psychometrically-aware framework that integrates diagnostic assessment with instructional scaffolding through a shared latent ability representation. PsyScore comprises three key modules: a Trait-Adaptive Neural IRT Scorer that incorporates the Graded Partial Credit Model (GPCM) into a neural architecture, enabling the precise estimation of student ability while maintaining psychometric interpretability, a ZPD-Scaffolded Feedback Generator, which conditions multi-agent feedback strategies on the diagnosed ability parameter to adapt instructional focus across different proficiency levels, and a Multi-Perspective Feedback Evaluation Strategy that assesses feedback quality via pairwise preference judgements and student revision simulations. Experiments on the ASAP++ dataset demonstrate that PsyScore achieves competitive scoring performance while providing more pedagogically aligned feedback.

Textual reuse pervades the Hebrew Bible, yet the computational methods used to detect it still rest largely on lexical overlap, and they falter once a parallel involves paraphrase, lexical substitution, or syntactic reworking. This paper introduces MiqraBERT, a Sentence-BERT model finetuned from AlephBERT (a Modern Hebrew encoder) for verse-level semantic similarity in Biblical Hebrew. The training set comprises 1,650 labeled verse and half-verse pairs: 825 true parallels drawn from the Chronicles synoptic material and from foundational studies of poetic parallelism, balanced against 825 randomly sampled negatives. Through cosine-similarity regression, the model learns an embedding space in which parallel verses cluster together and unrelated verses move apart. We evaluate separation with distribution-based metrics, Wasserstein distance and the overlap coefficient, across ten random seeds. MiqraBERT improves distributional separation 2.7-fold over the pre-trained baseline and reduces the ambiguous overlap region from roughly 24% to about 6%. Narrative synoptic parallels reach a recall@10 of 87.1%; poetic parallels remain difficult, below 9%. This genre-dependent asymmetry confines the model's reliable scope to narrative textual reuse. MiqraBERT is publicly available at https://huggingface.co/davidmsmiley/MiqraBERT

Structural ambiguity arises when a single sentence admits multiple valid interpretations due to its syntactic structure, posing a fundamental challenge for language understanding. Visual scenes serve as useful cues for resolving such ambiguity, and Vision and Language Models (VLMs) need to be capable of deriving possible semantic interpretations from visual scenes. We introduce Language and Vision Structural Ambiguity (LaViSA), a benchmark designed to evaluate the ability of VLMs to resolve structural ambiguity leveraging visual scenes. LaViSA consists of ambiguous sentences, their disambiguated sentences, and corresponding images of these disambiguated sentences across seven ambiguity categories. Using LaViSA, we conduct a comprehensive evaluation of diverse VLMs, including both proprietary and open-source models with varying parameter scales and reasoning capabilities. Experimental results show that although recent VLMs can leverage visual scenes to resolve structural ambiguity to a some extent, they still struggle with certain ambiguity types and visually subtle semantic distinctions, indicating remaining limitations in resolving structural ambiguity using visual scenes.

Real-world clinical decision support requires reasoning over heterogeneous and longitudinal patient information rather than answering isolated medical questions. However, current medical large language models and retrieval-augmented generation systems often rely on single-step prompting or retrieval, which can be fragile when clinical evidence is distributed across long electronic health records, medical images, sensor streams, guidelines, and referral constraints. This paper proposes MedRLM, a Recursive Multimodal Health Intelligence framework for long-context clinical reasoning, sensor-guided screening, and community-to-tertiary referral support. Instead of compressing all patient information into one prompt, MedRLM treats the patient case as an external clinical environment that can be recursively inspected, decomposed, retrieved, verified, and synthesized. The framework coordinates specialized agents for clinical text, longitudinal EHR, medical imaging, physiological sensor signals, guideline retrieval, uncertainty auditing, and referral planning. It further introduces a Clinical Evidence Graph Memory to connect patient-specific observations with retrieved evidence, standardized definitions, sensor-derived biomarkers, and referral criteria. A sensor-guided recursive triggering mechanism activates deeper reasoning when abnormal physiological or behavioral patterns are detected, while uncertainty-gated refinement supports clinician review for high-risk or low-confidence cases. We also outline a real-data evaluation design using public and credentialed clinical datasets spanning EHR, radiology, ECG, ICU time series, and referral-proxy outcomes. MedRLM aims to move medical AI from static question answering toward auditable, multimodal, and workflow-aware clinical decision support.

Training automated pronunciation assessment often relies on labeled learner errors or non-native corpora that are costly to collect. We propose a lightweight framework trained only on native speech resources, operating unsupervised or lightly calibrated with a small set of scored utterances. At inference, learner speech is discretized with an SSL encoder and a K-means codebook. A token language model trained on native sequences computes surprisal where higher surprisal indicates phonotactic deviation. We add a transcript-guided Text2DUnit--DTW module that predicts native token sequences from reference text and aligns them to acoustic tokens to derive error-sensitive features. Surprisal and alignment features are fused via simple regression. On SpeechOcean762, PCC improves from 0.60 to 0.66 with transcript guidance, near supervised baselines. Cross-dataset evaluation on L2-ARCTIC shows consistent gains.

Grapheme-to-phoneme (G2P) conversion for Modern Hebrew is needed for applications like text-to-speech (TTS), but is challenging due to the language's abjad writing system, which leaves vowels largely unwritten, creating substantial ambiguity. Standard approaches first predict vowel diacritics (nikud) to produce International Phonetic Alphabet (IPA) transcriptions, but this is limited: vocalization data is scarce and laborious to produce, it does not specify features such as lexical stress, and it reflects formal grammatical rules rather than everyday spoken pronunciation. Direct sequence-to-sequence IPA prediction, meanwhile, struggles on limited data and fails to exploit the character-level alignment characteristic of abjads. Our method, ReNikud, overcomes these limitations with two key insights: (1) Weak audio supervision via a phoneme-based automatic speech recognition (ASR) pseudo-labeling pipeline on thousands of hours of unlabeled Hebrew audio, yielding phonemic transcriptions that reflect natural spoken norms without manual annotation. (2) A pseudo-vocalization architecture that predicts IPA phonemes at each character position, enforcing character-level alignment as an inductive bias. Results on existing Hebrew G2P benchmarks and the new targeted MILIM benchmark for spoken Hebrew show that ReNikud surpasses previous state-of-the-art methods. We will release our code and trained models to support further work on Hebrew TTS and speech technologies.

Sign languages are expressive visual languages used by Deaf and Hard-of-Hearing (DHH) communities. Despite substantial progress in sign-language recognition, translation, and production, advances remain constrained by fragmented datasets, inconsistent annotations, and limited linguistic coverage. Existing benchmarks often fail to reflect real-world communication needs, and systematic analyses of these limitations remain limited. In this survey, we present a comprehensive index of sign-language datasets, covering 120 resources across 35 sign languages. We analyze key challenges such as modality imbalance, annotation granularity, and signer bias, and outline considerations for future dataset design. We also introduce a 24-field Sign-Language Datasheet and release a public GitHub repository (https://github.com/Ginqwerty/Open-Sign-Language) to support standardized documentation and reproducible evaluation. Overall, our work provides a unified and practical foundation for developing inclusive, robust, and scalable sign-language technologies in real-world applications.

The narrative composition of web-scale LLM pretraining corpora remains largely unexplored even though narrative is a fundamental mode of human communication. We present the first fine-grained study of narrative features in Dolma, a 3-trillion-token open pretraining corpus. Drawing on narrative theory, we design a framework spanning three core narrative elements (agency, setting, and events) operationalized as 11 interpretable dimensions. After sampling and annotating a diverse set of 400 passages, we finetune and validate NarraBERT, a RoBERTa-based model for fine-grained narrative prediction. We apply NarraBERT to 3M passages, resulting in a new dataset, NarraDolma. We find (i) narrative structure is measurable at scale across extremely heterogeneous data, (ii) we uncover a continuous, multidimensional narrative structure underlying web text, and (iii) narrative qualities are unequally distributed across pretraining sources and topics in ways that current curation practices neither measure nor account for. Our framework, dataset, and analyses provide a foundation for understanding how narrative qualities are distributed in LLM pretraining data and for studying how data composition affects narrative reasoning tasks. We publicly release NarraDolma and NarraBERT.

LLM-as-a-Judge has become the dominant evaluation paradigm for language models, but judge validation in practice relies on exact-match agreement, a metric that does not correct for chance and systematically overstates discriminative ability. We present the largest systematic evaluation of LLM-as-a-Judge to date: 21 judges from nine providers across MT-Bench, JudgeBench, and RewardBench, evaluated under three protocols (agreement, consistency, bias audit) over 118 runs and approximately 541,000 individual judgments. Four findings emerge, consistent across the full cohort, including the April 2026 frontier: kappa deflation between exact match and Cohen's kappa is universal (33--41 pp on MT-Bench), judge rankings shift by up to 14 positions across benchmarks, high test--retest reliability (>0.95) coexists with severe position bias (>0.10) in two production-deployed judges (instantiating a consistency--bias paradox), and verbosity bias is small (<0.011) across our cohort under a single pairwise rubric. We distill these into a Minimum Viable Validation Protocol.

Clinical NLP increasingly relies on electronic health record (EHR) data to detect suicidal behaviors, treating clinical documentation as more reliable ground truth than social media. We argue that this framing obscures how EHR-based suicidality datasets encode a particular operationalization of suicidality, shaped by who authors the data, how episodes are bounded, and how ambiguity is resolved. We ground this argument in a case study of the ScAN dataset, built over MIMIC-III clinical notes. We show how governance constraints, ICD-based cohort selection, single-annotator labeling, and hospital-stay-level aggregation produce labels that reflect clinician-documented judgments, treat suicidality as a bounded episode, and assume that intent can be reliably inferred from documentation. A linguistic analysis demonstrates that identical labels subsume heterogeneous clinical framings differing in temporality, negation, and uncertainty. We argue that clinical NLP should examine the assumptions embedded in suicidality datasets before interpreting their labels as ground truth.

Most companies read their customer support data at scale using sentiment analysis, which measures how customers sound rather than whether they were satisfied with the result. We tested a richer alternative on 70,450 support conversations from a leading online fundraising platform: alongside tone, we used GPT-5.4 to estimate each customer's satisfaction and to flag whether they reported a concrete problem, then validated all three readings against the 1-to-5 ratings customers left on the conversations they rated. The satisfaction estimate tracked those ratings far better than sentiment did, correlating at 0.47 against 0.36 and flagging unhappy customers with far fewer false alarms. The structured read also sees what sentiment cannot: tone and satisfaction disagree in 44% of conversations, a single "Neutral" label hides everything from quietly satisfied customers to ones who quietly gave up, and the largest group of all is "tolerated friction," customers who are satisfied but still reporting a fixable problem, a standing issue that no sentiment-based dashboard can surface. The broader finding is that LLM-based annotation can capture far more than the tonality of a customer's language, offering strong potential for new business metrics grounded instead in the customer's state (whether they were satisfied) and the cause of their problem extracted directly from the raw textual data of interactions and feedback.

Language models have become essential tools in shaping modern workflows. However, their global effectiveness hinges on a nuanced understanding of local socio-cultural contexts. To address this gap, we present NRITYAM, a comprehensive benchmark for evaluating the cultural comprehension capabilities of language models in the context of global dance traditions. NRITYAM comprises 9,260 carefully curated question-answer pairs spanning 12 languages, making it the largest dataset dedicated to evaluating cultural knowledge in dance. The dataset has been developed from the ground up through close collaboration with native dance artists and native speakers of the languages, who authored and validated culturally relevant questions specific to their regions. We evaluate a broad set of models, including large language models, small language models, multimodal large language models, and small multimodal language models. As a multilingual and multicultural benchmark, NRITYAM sets a new standard for evaluating the ability of AI systems to understand and reason about traditional performing arts. Detailed dataset samples are available at~\url{https://github.com/niladrighosh03/NRITYAM}.

Decomposing compound sentences into atomic, verifiable claims is a prerequisite for reliable automated fact-checking. Prior work has relied on token-overlap (Jaccard) metrics that systematically underestimate decomposition quality for paraphrastic claims, and has lacked formal termination analysis for the repair loop. We present Credence, a revised claim decomposition and evaluation framework addressing both shortcomings. Our contributions are: (1) Semantic-F1: we use BGE-large cosine similarity fidelity metric that resolves Jaccard's penalisation and improves downstream fact-checking accuracy; (2) Convergence theorems: we formally characterise four properties of the repair pipeline, establishing that rule-based repair is monotone and finitely terminating under an oracle parser assumption; LLM-based self-repair is provably non-monotone and requires an early-exit guard; (3) Three evaluation benchmarks spanning social-media, encyclopaedic, and news domains for cross-domain generalisation measurement; (4) Multi-model benchmarking across four decomposer models (3.8B-12B) and a closed API model. Experiments on SocialClaimSplit, WikiSplitBench, and ClaimDecompBench show that Semantic-F1 outperforms Jaccard-F1 by +15-32pp. EPR ranges from 0.94 to 1.00 on SocialClaimSplit and WikiSplitBench, while ClaimDecompBench includes lower base EPR cases (down to 0.824) due to harder news-domain constructions, and rule-repair reduces the Atomicity Violation Rate (AVR) by 47-100% relative to the base model without degrading fidelity.

Benchmark infrastructure for personally identifiable information (PII) detection remains limited: existing corpora cover few entity types, use ad hoc generation conditions, and do not show which surface conditions cause detector failures. We present REDACT, a systematically controlled multilingual PII benchmark with 13,427 records, 324,078 entity annotations, 51 entity types, 4,127 surface-form patterns, and 25 languages across 9 scripts. A strength-2 covering-array sampler controls nine generation axes: domain, format, difficulty, length, density, code-switching, language, adjacency, and co-occurrence. Three entity-level metadata fields (disclosure status, disclosure form, and a GDPR-aligned sensitivity tier) enable stratified evaluation beyond aggregate or per-type F1. From the full benchmark, we evaluate five detectors (Presidio, GLiNER, the OpenAI Privacy Filter, GPT-4.1, and Claude Sonnet 4.6) on a locked, language-stratified sample of 1,000 records. Aggregate F1 masks an architecture-dependent failure structure: the rule-based detector performs poorly on the highest-stakes data, including HIGH-sensitivity categories (recall 0.07) and non-verbatim disclosure forms, while the LLM detectors remain more robust, with the HIGH tier as their strongest sensitivity slice. A three-model reference-free LLM-as-judge assessment corroborates that sensitivity-tier assignment is the task's hardest axis. We release the benchmark, schema, prompts, and stratified evaluation harness.

Recent advances in Large Language Models (LLMs) have substantially transformed Automated Essay Scoring (AES), yet the internal mechanisms underlying LLM-based scoring remain poorly understood. In this work, we systematically analyze the hidden representations of eight LLMs across two English essay datasets (ASAP++, CSEE) and one Portuguese dataset (ENEM). Using linear probing, cross-prompt generalization, dimensionality reduction, and neuron-level analyses, we find consistent evidence that essay quality information is encoded in a linearly accessible form within LLM representations. These representations emerge progressively across layers, remain robust across prompting strategies, and partially transfer across essay prompts despite differences in scoring rubrics. In addition, nonlinear probes provide only marginal and inconsistent improvements over linear probes, suggesting that most essay quality information is already linearly decodable. We further identify individual ``essay scoring neurons'' whose activations strongly correlate with essay scores and whose behavior is sensitive to targeted intervention. Moreover, the layer-wise distribution of these neurons systematically shifts with essay length, with longer essays relying more heavily on deeper layers. Overall, our findings provide evidence that LLMs encode structured representations related to essay quality and offer new insights into the interpretability of LLM-based AES systems.

We present CzechDocs, a multiway parallel dataset of formatted documents (HTML, DOCX, and PDF) covering Czech and minority languages used in Czechia-primarily Ukrainian and English, with smaller portions of Vietnamese, Russian and other languages. The dataset is designed to support the evaluation of machine translation systems that aim to preserve document formatting during translation. We provide a comparison of the most common approaches to format-preserving machine translation on a validation subset of the dataset. This validation split, together with the evaluation toolkit, is publicly released for further research. A held-out test split will be reserved for a future shared task focused on document-level translation with formatting preservation.

We introduce the Meaning Intelligence Framework (MIF), a nine-dimension annotation and evaluation schema for Nigerian public discourse that separates surface sentiment from true communicative intent. Existing benchmarks for Nigerian languages, including NaijaSenti and AfriSenti, treat sentiment classification as a three-way polarity task (positive, negative, neutral). We argue that the dominant failure mode of AI systems on Nigerian discourse is not translation failure but context failure: the same utterance carries opposite pragmatic force depending on speaker, audience, and situation. The MIF operationalises this insight across nine scored dimensions: register, surface sentiment, true intent, irony, coded subtext, risk tier, annotator confidence, speaker emotion, and recommended communications action. We construct a 30-item calibration dataset spanning Standard English, Nigerian English, Nigerian Pidgin, and code-mixed registers, and evaluate a frontier language model (Gemini 2.5 Flash) under zero-shot and schema-informed prompting conditions. The headline finding is the Register Gap: zero-shot register classification accuracy is 33.3%, rising to 73.3% (+40 points) when the model receives the MIF schema in-context. The composite Meaning Intelligence Score increases by 5.4 points (73.2 to 78.6) under schema-informed prompting, with the largest practical gains in register identification, coded-subtext detection (+10 points), and strategic action recommendation (+10.3 points). We release the framework specification, annotation guidelines, and the 30-item public calibration set to support reproducibility, while retaining a private holdout corpus for contamination-protected evaluation.

Online hate speech and misinformation frequently overlap, yet NLP research has mainly treated them in isolation. While LLMs represent a scalable solution for assisting humans in the generation of counterspeech for both threats, zero-shot models frequently generate repetitive and vague responses, underscoring the need for high-quality examples to steer model generation. However, existing counterspeech datasets against the overlap of hate and misinformation are scarce and limited to single-turn English dialogues, while real-life interactions span across multiple turns and languages. To bridge this gap, we introduce the first large-scale, expert-curated, multilingual dataset of dialogues tackling the intersection of hate and misinformation. To ensure factual grounding, the dialogues are also anchored in verified external knowledge (i.e., fact-checking articles and NGO reports) and include document- and chunk-level span annotations, making it directly applicable for RAG systems. Covering five languages and targeting hate directed at seven marginalized groups, this novel resource enables the training and evaluation of more persuasive, factually grounded counterspeech models.

Large Language Models (LLMs) exhibit representational and syntactic biases that are difficult to evaluate due to the stochastic nature of text generation. Standard auditing methods rely on a single output inspection or static automated metrics. These approaches obscure the underlying probability distributions and fail to capture biases hidden in lower-probability generation branches. This paper introduces TreeTracer, a visual analytics tool designed to evaluate LLM bias through aggregated comparison. Using a systematic perturbation analysis pipeline, the tool replaces ontology-defined terms in each input prompt, aggregates hundreds of stochastic generations into a syntax-aligned hierarchical structure, and then performs classification-aware node merging with an auxiliary language model. The resulting structure is visualized through a custom Sankey diagram. By juxtaposing two ontology-driven trees, the workspace enables direct comparison between semantic contexts and supports systematic bias detection. Because any visualization reflects only a subset of the model's learned behavior, the system further applies contrastive inference to compute and directly display counterfactual token probabilities across contexts, reducing the risk of misinterpreting the presence of bias. We validate the workspace through case studies comparing an unaligned baseline model GPT-2 XL against the constitutionally aligned Apertus models. The visual aggregation successfully exposes hidden representational harms, such as counterfactual pronoun suppression and conversational marginalization of individuals. A preliminary user study confirms that the aggregated comparative interface reduces cognitive load and effectively supports analysts in detecting systemic biases.

Large language models (LLMs) increasingly review and revise text, including their own. A documented self-preference bias (models favoring their own generations when acting as judges) raises the question of whether models also resist valid corrections to their own writing. We test this in a setting where "valid" is decided not by another model but by a deterministic verifier: instruction-following revision on IFEval. A model writes a draft; the official IFEval checker confirms the draft violates a constraint and that a candidate edit fixes it; the model then accepts or rejects that edit either as the genuine in-context author or as a fresh model that sees the draft neutrally. Across four mid-tier model families and 85 author-versus-fresh comparisons, we find no detectable self-preference: authors reject verified-good fixes to their own drafts at essentially the same rate as fresh models judging the same drafts (gap -5.1 pp, 95% CI [-12.9, +2.7]). A self-skepticism hint from a smaller pilot did not replicate at scale. The one robust observation is qualitative: when authors do reject a verified-good fix, 97% of their stated reasons are flaw-catching rather than preference, that is, about the character of rejections, not an elevated rate. Effects smaller than ~13 pp cannot be excluded at this sample size.

Fine-tuning language models on insecure code induces emergent misalignment with poorly understood internal structure. We investigate whether this misalignment corresponds to a causally actionable activation-space direction shared across architectures. Across four instruction-tuned model families (Qwen2.5-1.5B, Gemma-2-2B, Llama-3.2-1B, Ministral-3-3B) finetuned identically, a difference-in-means direction achieves 99.6% separation of aligned and misaligned activations at each model's final layer. Causal steering by subtracting this direction reduces code spillover by 21-51 points, while a secure-code control confirms content specificity. Cross-architecture transfer via ridge regression maps yields large behavioral suppression (up to 46 points) but fails specificity controls as random and orthogonal directions perform comparably. We identify a two-tier specificity structure: within-model directions are causally specific and actionable; cross-model directions are causally real but non-specific. An asymmetric transfer topology emerges, with Gemma and Qwen acting as geometric donors and Llama as a receiver. These findings define the limits of linear cross-architecture correction and recommend within-model probing for auditing.

Multimodal large language models (MLLMs) are increasingly deployed in personally and societally consequential settings, yet the visual cues that shape how these models judge people remain poorly understood. Prior work often compares different (groups of) individuals, making it difficult to separate appearance effects from identity differences. We introduce StylisticBias, a controlled benchmark for evaluating attribute-level social bias in MLLMs. We generate 500 photorealistic base faces and create about 50 single-attribute variations per face, producing about 25K images. This design keeps identity fixed and changes one visual attribute at a time. It lets us measure how specific cues shift model judgments. We evaluate six MLLMs across 25 binary social judgment scenarios. We find that age and body type dominate identity-level effects, while fashion style and other visual cues drive the largest attribute-level shifts. We further find that about 15 attributes account for nearly 80\% of the total variation, showing that bias is concentrated in a small set of visual cues. Sensitivity is strongest in judgments that are semantically aligned with appearance, especially socioeconomic and style-related judgments. We release StylisticBias as a benchmark for fine-grained bias evaluation in multimodal models. Code and dataset: https://github.com/timo-cavelius/StylisticBias and https://hf.co/datasets/shaghayegh/stylistic-bias-dataset.

Translating sequential programming priors into the parallel temporal logic of hardware design remains a crucial bottleneck for large language models(LLM). To investigate this, we introduce a new error taxonomy grounded in problem solvability, inspired by cognitive theory. Our taxonomy categorizes failures into syntactic, semantic, solvable functional, and unsolvable functional types. Evaluations reveal a strict empirical ceiling on the VerilogEval benchmark, as frontier models plateau at a 90.8% initial pass rate. These plateaus are defined by unsolvable functional errors, exposing persistent knowledge gaps immune to test time compute scaling. Furthermore, we expose a striking surface convergence gap: optimization readily eliminates syntax errors but concurrently exacerbates deeper functional failures. Our findings demonstrate that alignment techniques merely teach models to compile. While repeated sampling strategies can patch solvable errors, register-transfer level(RTL) coding capacity remains strictly bounded by pretraining knowledge. Addressing challenges in the current LLM based hardware generation pipeline requires more studies in model reasoning rather than alignment interventions.

We present a multilingual computational discourse analysis of how language constructed the algorithmic consecration of Vozinha, the 40-year-old Cape Verde goalkeeper, after Spain 0-0 Cape Verde at the 2026 FIFA World Cup. The study contributes a multilingual corpus in Portuguese, Spanish, English, and French; a nine-frame narrative taxonomy with cue-based frame annotation; a reproducible annotation pipeline combining LLM-assisted suggestion with human validation; and an analysis of cross-lingual narrative diffusion across discourse phases. We treat the platform follower count itself, narrated as "50k to 8M", as a linguistic object: a circulating and narratable proof of visibility rather than a mere measurement. The follower-growth timeline is used only as contextual metadata: we reconstruct a conservative phase structure, not a continuous API-native series, and type every datapoint by value class, confidence, and evidence type. The only exact primary scraper anchor is 8,235,652 followers at 2026-06-16 15:47 UTC; all other figures are reported as estimated ranges or thresholds, including an estimated pre-match baseline of 45k-56k. Findings suggest that distinct languages carried distinct frames: Portuguese mobilization, Spanish crisis, English nation-making, and a shared platform-metric spectacle through which peripheral athletic performance became globally visible. As a v0.1 pilot, the paper releases the corpus schema, frame taxonomy, annotation guidelines, hashed visual-evidence log, and typed timeline, while flagging full double annotation and inter-annotator agreement as planned work.

The increasing prominence of Large Language Models (LLMs) in public discourse presents both opportunities and challenges for democratic deliberation. While red teaming strategies help mitigate specific risks, broader concerns persist regarding linguistic constraints, biases, and the sycophantic tendencies of LLMs. This chapter explores how LLMs can be used to significantly scale up and democratise deliberation, particularly in fostering inclusivity and empowering traditionally marginalised groups. Drawing on concepts from Systemic-Functional Linguistics, the chapter examines how variations across language users (for example, with respect to socio-demographic groups) and across language use (for example, with respect to communicative functions) shape participation in AI-supported deliberation. The chapter presents AI-driven deliberation studies and assesses their potential to scaffold argumentation, enhance access, and reduce the influence of exclusionary linguistic norms and biases which are embedded in prestigious registers. At the same time, the chapter cautions against both overclaiming, which leads to unrealistic expectations, and underclaiming, which risks missed opportunities for AI-assisted engagement. The chapter concludes by identifying future research directions to maximise the democratic potential of AI-assisted participation while embedding ethical safeguards to counteract the reproduction of linguistic inequalities.

We present an algorithm for pitch spelling and key estimation. Given an input in MIDI-like format, containing information on note pitches (expressed in semitones relative to the lowest reference note) and bar boundaries, it estimates the appropriate note names, a global Key Signature, and a local scale for each bar. This related information elements are evaluated jointly during two stages of optimisation. During an initial 'modal' stage, a probable scale is proposed for each bar, minimising the number of accidentals to be printed in the printed score with a shortest-path search. Then, during a second stage called 'tonal', these local scales are used to estimate the Key Signature and note names that would result in the best musical notation for the entire piece. We present evaluations conducted on datasets comprising a variety of digital musical scores: jazz lead sheets taken from the Real Book, transcriptions of recordings of jazz soli and bass lines, traditional tunes, as well as classical scores for piano and monophonic instruments. Our procedure was originally designed for use in music transcription, specifically for building digital collections of jazz solos transcribed from audio recordings, for the purposes of music analysis, teaching and the preservation of cultural heritage. This method should also prove useful for other tasks related to the processing of musical notation. Furthermore, to this end, we have defined new distances between various common jazz scales, which may be of some interest to musicological studies.