Papers

Self-harm content is particularly challenging to detect using NLP techniques, and is also a high-stakes task which requires the highest accuracy to enable timely intervention or flagging at-risk users. We therefore present an analysis of how LLMs represent such self-harm content, which has downstream applications in self-harm detection, LLM intervention and governance and policing. In this paper, we focus on two datasets and four models, and perform two main experiments: (1) We train and evaluate linear probes across all layers of each model on two self-harm datasets: X-Sensitive and SH-Detection. Across both corpora, self-harm information crystallizes in the final 3 - 7% of network layers (93 to 97% depth). (2) We extract contrastive self-harm directions and, after performing a normaliation step, we find that the most accurate probes are not necessarily the most linearly separable. In particular, we find Gemma-3-4B to represent this contrastive self-harm direction in a slightly different, more intricate way than the other LLMs.

Chain-of-thought prompting improves language-model reasoning by carrying intermediate states across successive computation steps. However, relying on natural language as the only recurrent interface is overly restrictive, since many transient computations do not need to be fully verbalized. Existing latent-reasoning methods remove this constraint by recurrently propagating continuous hidden states. However, these methods pass a dense hidden vector as a whole, without an explicit mechanism for selecting and organizing the information needed by the next reasoning step. This motivates an intermediate interface that remains linguistically grounded without requiring a decoded sentence. We introduce J-CoT, a recurrent reasoning framework built on J-space, a vocabulary-indexed coordinate system within the model's hidden representations. Within each cycle, the model computes in its full hidden space. At the cycle boundary, J-CoT expresses the intermediate state as vocabulary-indexed coefficients, carries these coefficients forward as a J-thought, and maps them back into the model's hidden representation for the next cycle. J-CoT therefore requires neither a fluent intermediate rationale nor recurrence over the complete hidden state. Under matched backbone and inference settings, J-CoT-Zero matches or exceeds the strongest evaluated latent-reasoning baseline on every benchmark, while J-CoT-Train obtains the highest score across the evaluated mathematical, scientific, coding, and structured path-reasoning tasks.

Mastodon as a decentralized federation of independently moderated social servers poses unique challenges for the detection and mitigation of toxic content. There are no unified moderation standards. The ecosystem is very diverse and uneven. This paper explores the development and spread of toxicity in Mastodon, utilizing machine learning methods to examine user posts. The results offer clarity on toxicity trends and its implications for community health and decentralized governance.

Mixture-of-Experts (MoE) architectures have been widely adopted in large language models, yet parameter-efficient fine-tuning (PEFT) for MoE models remains underexplored. Existing PEFT methods for MoE either ignore router priors with uniform adapters, reducing efficiency and risking forgetting, or rely on static expert selection, limiting per-token capacity and cross-expert feature learning. In this paper, we make the first attempt to fine-tune MoE models with MoE-style low-rank adaptation: our method, entitled MoE$^2$-LoRA, deeply couples the pretrained expert specialization with task-specific adaptivity via a dual-channel Routing-Conditioned Projection (RCP) module, which reuses base router activations to inform LoRA routing. We further introduce a single global LoRA expert pool shared across all layers, enabling model-wide adaptation with emergent layer-wise affinities and balanced expert utilization. MoE$^2$-LoRA simultaneously benefits from the advantages of prior reuse, dynamic adapter routing, and model-wide knowledge sharing. Evaluated on multiple MoE backbones with varying scales and expert granularities, MoE$^2$-LoRA consistently achieves state-of-the-art downstream accuracy while retaining stronger general capabilities.

Test-time scaling through iterative self-evolution with environment feedback, as demonstrated by AlphaEvolve, shows remarkable performance gains. We hypothesize that the success of such evolution frameworks hinges on meta-skills, such as self-reflection with environment feedback, that enable effective multi-round refinement, yet are largely neglected by traditional post-training. To bridge this gap, we present MetaEvolve, a framework designed to develop these meta-skills via a data synthesis pipeline, evolution-aware reinforcement learning (RL), and inference-time evolutionary search. Concretely, we ground MetaEvolve in coding, where program execution provides natural, continuous reward signals beyond binary correctness. Building on these signals, we synthesize evolution trajectories as training data, each containing a current program, its fitness score (combining correctness and efficiency), and a history of prior attempts, and train the model via RL with verifiable rewards derived from test case execution. By training on large-scale code data, we aim to inspire generalizable domain-agnostic meta-skills that can transfer broadly to open-ended problems where such rich training signals are scarce. Across seven coding benchmarks, MetaEvolve outperforms the strongest baseline by 10.01% absolute on in-distribution tasks and 24.12% on out-of-distribution tasks. On open-ended algorithm optimization problems entirely outside the training domain, it further achieves a 46.9% relative improvement. These results demonstrate that explicitly cultivating self-evolution meta-skills offers a principled path toward more capable and autonomously self-evolving AI.

Benchmarks for LLM-agent memory typically generate conversations first and extract answer keys afterwards – with documented label-error and contamination problems – and they overwhelmingly measure short interaction histories. We invert the pipeline: a seeded life-script sampler emits facts with validity intervals, volatility classes, and source channels before any text exists; an LLM renderer writes chat and email from per-event fact manifests; a fidelity verifier confirms every planted fact; and questions are instantiated mechanically from the script, so gold answers are script-valid by construction and separately validated for answerability. The synthetic, fictionalized corpus (~380 questions, 15 types) embeds features absent from the benchmarks we survey: per-fact validity intervals, sent/received trust distinctions, injection probes in a benign harness, and as-of-date question sets. Benchmarking five memory architectures against a no-memory control (fixed answerer, versioned LLM judge, three replicates, two horizons), we find backend rankings invert with history length: the budgeted curated-map memory that leads at three weeks loses recall of evicted content by nine weeks (96% to 72%) while a provenance-typed graph rises to 90%; the inversion is positive for all six users under complete cross-family re-judging (exact p=0.031). A full-rendered-history baseline ties or exceeds the best memory system at the short horizon but shows no judge-independent advantage at nine weeks, at about twice the read cost. Write-stage quality strongly correlates with downstream quality (weakly-written facts fail 24% vs 2%), and injection resistance tracked whether provenance boundaries survive representation. A layered architecture performs best among the memory systems in both regimes (96.8% short-horizon) and is released as Veracium, an open-source library, with the corpus generator and harness.

Large language models (LLMs) are increasingly used to generate long-form conversational content such as podcasts from textual sources. While these systems produce fluent and engaging narratives, they often introduce ungrounded information. In this work, we present the first systematic study of faithfulness in document-grounded podcast generation, where grounding must be maintained across conversational turns in long-form, multi-speaker transcripts. We construct a dataset of over 1500 documents spanning five domains and generate podcast transcripts using multiple LLMs. We introduce a turn-level LLM-as-a-judge framework for evaluating whether conversational turns are supported by the source document, and validate its reliability through human studies. Our analysis shows that even state-of-the-art models, including GPT-4o, frequently generate ungrounded content. To mitigate this issue, we propose catch-n-repair, a model-agnostic framework that detects and rewrites unfaithful conversational turns while preserving conversational flow. Experiments demonstrate consistent improvements in faithfulness across both in-domain and out-of-domain settings.

Historical documents act as invaluable knowledge archives but often suffer from illegibility due to physical deterioration and damage. While existing restoration methods based on masked language modeling effectively utilize local context, they struggle to restore named entities that require external historical knowledge. To address this limitation, we introduce a novel framework for historical document restoration that leverages large language models with retrieval-augmented generation (RAG). By combining the implicit knowledge of pre-trained LLMs with explicitly retrieved external context, our model ARI effectively mitigates the challenge of inferring context-dependent proper nouns. Extensive experiments on Korean historical documents demonstrate that our approach significantly outperforms baselines, achieving substantial gains in restoring both general characters and named entities. Furthermore, comprehensive evaluations including expert assessments confirm that ARI serves as a practical tool for domain experts, promising to accelerate the analysis of historical records.

Image inpainting aims to reconstruct missing or corrupted regions of an image while preserving as much as possible, visual and semantic consistency. In medical imaging, this task is particularly important because artifacts, missing information, and pathological alterations can compromise diagnostic reliability and downstream clinical applications. Recently, diffusion models have emerged as state-of-the-art generative approaches for medical image inpainting due to their ability to generate anatomically consistent reconstructions. This survey presents a systematic review of diffusion-based methods for medical image inpainting, covering the main architectures, applications, datasets, and evaluation strategies reported across 60 studies. In addition, we propose a taxonomy for diffusion-based approaches. The analysis reveals a rapid growth of research interest in diffusion-based medical image inpainting, with denoising diffusion probabilistic models and latent diffusion models emerging as the dominant architectures. The reviewed studies mainly focus on artifact removal, data augmentation, pseudo-healthy tissue reconstruction, and anomaly detection, particularly in magnetic resonance imaging and computed tomography imaging. Overall, diffusion models demonstrate strong performance in producing anatomically plausible reconstructions and aiding downstream clinical tasks. However, the review also highlights important challenges, including the lack of standardized benchmarks, limited dataset diversity, and restricted validation procedures across diverse clinical applications and imaging scenarios.

Foreign language anxiety (FLA) can be a major barrier to second language acquisition (SLA), especially in conversational contexts. With the proliferation of large language models (LLMs) throughout all areas of life, recent work suggests that interacting with LLM agents can be instrumental within the field of SLA and foreign language education, especially for reducing FLA. Related work also suggests that linguistic demands and task complexity can be predictors of FLA, implying that the use of demanding, complex language could lead to learners experiencing higher FLA. In this paper, we propose a novel multi-agent embodied conversational system that generates level-appropriate dialogue for English language learners. These levels are based on those defined by the Common European Framework of Reference for Languages (CEFR) to describe non-native listener and speaker proficiency. Using a "generate-evaluate-regenerate" loop with multiple LLM agents and a level classifier, it achieves a desired simplicity that is adaptive to the user's proficiency level. We also share the results of a preliminary small-sample pilot study that tested this system with Japanese university students, to see whether it would yield lower FLA levels than an unsimplified embodied conversational agent. Analysis of conversational output showed that 87.4% of dialogue sentences generated by the proposed multi-agent system fell within one predicted CEFR level of the learner's self-assessed proficiency, compared to 54.1% for the unsimplified agent. This suggests that the novel system is better able to produce output at an appropriate level for the learner. Though this study did not yield statistically significant evidence that the system reduces FLA levels in Japanese learners of English, likely due to a small sample size, it provides usability findings and culturally-informed design insights that will inform future study.

An LLM producing the response pattern associated with a human psychological effect is not the same claim as the LLM possessing that bias. We present PsyAgentBench, a benchmark that re-runs classic psychology experiments on LLM agents under a factorial design built to separate these: each paradigm is run with the paradigm explicitly labeled in the prompt (named) or framed as a routine task (blind), and on the literal textbook version of the task (canonical) or a structurally matched variant written to reduce lexical and scenario overlap with likely training data (counterfactual), crossed with a persona manipulation. Across five completed paradigms, evaluated on up to three open-weight model families with 41,904 trials released, apparently human-like effects arise through qualitatively different routes rather than one susceptibility: paradigm-label gating with explicit override (Asch conformity, 0 percent blind to 83.3 percent named on gpt-oss-120B), knowledge-dependent signal reliance (anchoring, exactly zero on grounded facts versus near total on invented quantities, a pattern equally consistent with rational use of the only available signal), amplification on novel content under labeling (framing), robust absence (sunk cost), and safety-mediated selection where refusal itself is the primary finding (minimal-group allocation). A one-sentence persona change (agreeableness, framed as an instruction rather than a verified trait manipulation) eliminates, dampens, or reverses these effects depending on which effect it is, arguing against any single response-bias account. We further formalize, and in two cases document empirically, three ways a psychology paradigm can fail to port to LLM agents: persona dominance, population collapse, and safety selection. We argue scalar bias-susceptibility scores obscure this structure and report replication profiles instead.

We introduce TatBLiMP, the first benchmark of linguistic minimal pairs for Tatar (tt, ISO 639-3 tat), a Qypchaq Turkic language written in Cyrillic. To our knowledge it is the first grammaticality evaluation for Tatar language models of any kind, since even the 101-language MultiBLiMP does not include Tatar. TatBLiMP covers 16 morphosyntactic phenomena in 1248 sentence pairs. Each pair differs by a single morpheme, one grammatical and one ungrammatical. A model passes a pair when it assigns higher probability to the grammatical member. Scoring compares probabilities the model already assigns, so the benchmark needs no text generation and no parser, and it runs on base models and on mid-training checkpoints. TatBLiMP adapts the phenomenon inventory and single-morpheme breaking operations of TurBLiMP to Tatar and adds one phenomenon specific to Tatar, bare-noun number after numerals and quantifiers. The grammatical member of every pair is an attested sentence from Tatar literary prose. The ungrammatical member is produced by a deterministic single-morpheme perturbation with the apertium-tat transducer. Every pair is ratified by a native speaker. A plausibility principle governs construction, so the ungrammatical member is a plausible real-world error rather than an arbitrary corruption. Across from-scratch Tatar models, cross-lingual adaptations, and frontier multilingual LLMs, the benchmark tracks focused Tatar training rather than parameter scale. A 478M from-scratch model and a 125M monolingual model lead near 0.97, a 7B adaptation trails, frontier LLMs of 30-120B parameters fall to 0.80-0.92, and a lightly tuned multilingual model is weakest. We close with the benchmark's main limitation. Its inherited taxonomy omits the morphophonology, vowel harmony and consonant assimilation, that is most salient to native speakers, and we sketch a native second layer that would add it.

We study when limiting what a language model can see improves learning. We compare standard CoT, the more general learner that reads the full trace, with recursive language models, which restricts itself by solving each subtask in an isolated context. In-distribution, this generality comes for free: CoT can efficiently simulate the recursive rule, so the IID generalization guarantee changes only by a constant factor, and recursion does not offer much. But out of domain, CoT can fit training by relying on context outside the current subtask, i.e. a shortcut that breaks once those tokens change; recursive context isolation rules out this failure mode. Even though CoT's class still covers the recursive rule, simplicity bias picks the shortcut over the truth. Thus, to go beyond distributional accuracy and truly reason, covering the right rule is not enough; this contrasts with classical learning theory.

Language models write text once, left to right, and models that write in other orders, such as diffusion LMs, pay for it with many full-sequence passes. We introduce Reviser, a decoder-only Transformer that writes and edits text with cursor actions (insert, move, delete, stop) that a deterministic executor applies to a canvas. Reviser is autoregressive over its own actions, so a step costs the same as an autoregressive (AR) step. We train Reviser to continue text, including a self-correction recipe that types a wrong token and deletes it, and Reviser-Edit to revise a draft, and compare each with a same-size AR model trained on the same data, using three LLM judges on continuations cut to the same length. Reviser-Edit beats its matched AR rewriter on CoEdIT SARI (+7.2 at 100M, +4.5 at 300M; two seeds agree) with a third of the decoding steps and latency, though the AR rewriter corrects grammar better. In continuation, Reviser beats diffusion LMs of similar size (62-74% win rate) at a fraction of their compute; two of three judges prefer its AR twin, and one slightly prefers Reviser at 100M. Reviser fills gaps it left and backspaces a just-typed repeated token with probability 0.85-0.87.

Next-Generation Sequencing has revolutionized the study of genetic mutations, enabling large-scale investigations into their roles in disease development. However, extracting meaningful insights from the vast amount of biomedical literature remains a complex challenge that cannot be addressed manually. In this paper, we present pre-trained models (PTMs) for the automatic extraction of relations from biomedical text, specifically targeting the variant-phenotype domain. Our evaluation on the SNPPhenA corpus demonstrates that fine-tuning small BERT-based models, particularly DeBERTa, yields strong performance, approaching the current state-of-the-art (SOTA). Additionally, our results indicate that carefully fine-tuning Google's Gemini Pro 1.0 outperforms the existing SOTA for both sentence-level tasks (where the model processes only the target sentence) and abstract-level tasks (where the model processes the entire abstract).

Extracting SNP-phenotype associations from biomedical literature is vital but challenging. We benchmarked diverse NLP models, including MLMs, hybrid architectures, and state-of-the-art LLMs (Gemini 2.0, OpenAI O-series, Qwen, Mistral), on the SNPPhenA corpus across three tasks: sentence-level, abstract-level, and association strength classification. OpenAI O1 achieved state-of-the-art (SOTA) results using few-shot learning for non-finetuned sentence-level classification (F1 0.89) and established a new SOTA for abstract-level classification (F1 0.82). Association strength classification proved difficult, though fine-tuned Gemini 2.0 Pro performed best (F1 0.60) in the first LLM evaluation of this task. Proprietary LLMs, especially in few-shot (O1) or fine-tuned (Gemini 2.0 Pro) settings, significantly outperformed other models. These findings confirm the power of modern LLMs for genomic knowledge extraction.

Large language model (LLM) personas are proposed as survey respondents, yet validation outside English-speaking contexts is scarce. We evaluate how well a Korean synthetic persona panel used to condition Gemini 3.5 Flash and EXAONE reproduces digital and artificial intelligence (AI) service-use distributions of the Korea Media Panel Survey. About 8,000 personas per model answered eight service-use items and eight attitudinal constructs; responses were compared with weighted survey estimates. The overall mean absolute error (MAE) was 14-19 percentage points (pp), with binary item-mean correlations of 0.70-0.91 across waves. Segment error across five axes was 14-18 pp, with between-group signed-error ranges of 49.6/34.7 pp (Gemini/EXAONE; 39.5/31.2 without the non-comparable teen cells). Errors were model-specific: an age stereotype (Gemini) versus an acquiescence-consistent level bias (EXAONE). Generative-AI overestimation was consistent with temporal misalignment; short-form underestimation was framing-sensitive and persisted under randomized order (both shown for Gemini). Post-hoc holdout calibration on 30% of the real data, with the correction form selected inside the calibration set, cut cell MAE from 18.3/15.1 to 4.9/4.4 pp, yet direct estimation from that subsample was more accurate than the calibrated panel (3.6 pp), a synthetic-informed shrinkage estimator beat its real-only counterpart by at most 0.7 pp, and the correction did not transfer competitively across waves. The calibrated panel kept an advantage only below roughly 250-860 real responses (at most 2.3 pp over a real-only shrinkage estimator) or, for one model, on unobserved segments. In this setting, synthetic panels are not survey substitutes; their value is diagnostic.

Sequence-to-edit approaches make grammatical error correction (GEC) efficient and locally interpretable by predicting edit labels over the input rather than generating a full corrected sentence. Their interpretability, however, is primarily operational: a label specifies how the string should change, but a single edit vocabulary does not always reveal the type of correction being made. We propose STAGEET, a stage-wise typed edit-tagging framework that reorganizes Seq2Edit supervision into typed executable stages and extends edit operations to correction categories. STAGEET decomposes correction into an ordered sequence of medium-grained typed stages; each stage predicts from its own label space, rewrites the current hypothesis once, and passes the resulting intermediate sentence to the next stage. We instantiate the framework as both an end-to-end shared-encoder multi-head model with stage-specific adapters and a fully specialized variant with one independent tagger per stage. Experiments on QALB-2014 and ZAEBUC show that category-aware staged correction retains competitive edit-based GEC performance while exposing a more inspectable correction trajectory, and attains state-of-the-art results on QALB-2014.

Domain adaptation of small language models (SLMs) has emerged as a practical strategy for deploying capable NLP systems in resource-constrained, high-stakes environments including healthcare, legal services, and financial analysis. While performance gains from parameter-efficient fine-tuning are well characterised, the corresponding impact on trustworthiness (factual calibration and adversarial robustness) remains poorly understood. This paper presents the first systematic cross-domain, cross-architecture empirical study quantifying the trustworthiness cost of domain adaptation across three SLM architectures (TinyLlama 1B, Gemma-2 2B, Llama 3.2 1B), three domains (healthcare, legal, finance), two training-data conditions (benign and adversarially perturbed), and four fine-tuning strategies (baseline LoRA, Safety-DPO, Dark Experience Replay, and Task Arithmetic LoRA, TA-LoRA). Trustworthiness is evaluated through TruthfulQA MC2 (factual calibration) and HarmBench ASR (adversarial robustness) across all 216 experimental configurations with three random seeds. Three principal findings emerge. First, baseline QLoRA domain adaptation produces minimal TruthfulQA MC2 change across all model-domain combinations (mean |Delta TQA| < 0.02). Second, adversarially perturbed training data consistently improves domain adaptation quality (Delta loss approximately -0.040) without worsening trustworthiness benchmarks. Third, none of the three safety-preserving strategies reduced adversarial harm susceptibility: Safety-DPO was effectively neutral (mean Delta ASR < 0.001), while Dark ER and TA-LoRA increased mean HarmBench ASR by +0.171 and +0.155 respectively in safety-aligned models (Gemma-2 2B, Llama 3.2 1B), with individual configurations exceeding +0.45. These results challenge the assumption that replay-based and arithmetic-merge strategies transfer alignment to domain-adapted SLMs.

We study baking documents directly into the weights of a 4-bit Gemma-4-e4b model via LoRA, so a system can answer questions about a corpus closed-book: no retrieval and no context-window budget. Across roughly 100 training runs from single documents to a 99-document corpus, we find that once adapter capacity is adequate, training-data quality is the dominant lever on closed-book accuracy, outweighing LoRA rank, learning rate, and two alternative architectures combined; capacity itself is a hard gate below which no data intervention helps. A single curation pass (shortening gold answers to canonical 1-6 word spans and dropping trivia) moved closed-book accuracy from 57.7% to 85.7% on a 15-document corpus, a larger jump than any architectural change. We confirm a capacity trend (rank must grow with corpus size) entangled with a coupling between rank and learning rate that we initially misdiagnosed. On a 15-document slice we add a real retrieval baseline: the internalized adapter (84.2% recall) beats a BM25-RAG pipeline with a base reader (58.9%) and even a realistic gold-chunk oracle (65.6%) at lower latency. We report the full arc, including three misdiagnoses, as a case study in debugging LLM training empirically.

Modern reasoning models depend on reasoning data, today sourced from human annotations or distilled from stronger LLMs. However, a rich and largely untapped source of supervision lies in expert systems (e.g., game engines, classical planners, theorem provers), which routinely produce near-optimal actions across diverse domains. But these experts are silent: they commit to an action without writing down the chain of thought (CoT) behind it. Recovering that CoT as natural-language reasoning would distill expert knowledge into a student that generalizes beyond the demonstrated actions. We treat it as a latent variable and study how to recover it from the action alone. Our approach, LeAct (Learning to reason from Actions), optimizes this latent variable: the student samples candidate CoTs for each expert action, and we retain those that measurably improve its own probability of recovering the action. Across imperfect-information games at multiple scales and a simulated robotics benchmark, LeAct reaches the solver's numerical floor on small enumerable games. At larger scale, it is $5\times$ closer to the solver than the strongest expert-iteration baseline. At Flop Hold'em ($\sim 10^9$ infosets), LeAct wins head-to-head by $+60$ mbb/g, and on the robotics probe it is the only training recipe that improves on direct imitation. We present a principled framework and the result: expert systems become a categorically new source of reasoning teachers for foundation models.

Lossless acceleration schemes, such as speculative decoding, promise significant inference speedups by relying on dynamic token-level alignment between a draft and a target model. However, this guarantee of semantic equivalence masks a severe operational vulnerability: draft-target alignment can be systematically attacked. In this paper, we introduce ADSD, which, to the best of our knowledge, is the first prompt-suffix attack that collapses verifier acceptance by pushing draft probability mass toward tokens the target is unlikely to accept. ADSD uses Soft-Collapse, a verifier-aligned surrogate derived from the asymmetric speculative acceptance rule, together with a target-preservation objective that discourages obvious task corruption. ADSD successfully generates highly effective adversarial suffixes. On the GSM8K dataset, our attack increases the mean sample time by 62.3% while preserving the task quality. We further show that this vulnerability exists across different domains, speculative decoding strategies, and model architectures.

Large language model (LLM) agents increasingly perform commercial tasks that involve retrieving external content, such as images, and, where appropriate, reproducing that content. LLM agents should comply with the law, including copyright law. Presently, however, we lack adequate frameworks to assess whether they do so in practice. To that end, we introduce Copyright-Bench, a benchmark designed to evaluate LLM agents' compliance with copyright law. Copyright-Bench comprises realistic commercial tasks—website development, merchandise design, and pitch deck production—that involve agents selecting between public-domain content, the use of which is legal, and copyrighted content, the use of which is infringing in this setting. The evaluation introduces prompt variations that simulate different user preferences, as well as time pressure. Comparing state-of-the-art LLM agents against a human baseline, we find that: (1) agents select copyrighted works despite the availability of public-domain alternatives; and (2) for open-weight models, violation rates increase in response to certain user preferences and simulated time pressure.

Roadblocks in Bolivia are a social conflict phenomenon with devastating economic impacts, estimated at losses equivalent to 4% of the national Gross Domestic Product. Despite their recurrence and impact, there is a lack of local predictive systems to anticipate these events for logistical decision-making. This paper presents a hybrid probabilistic forecasting system that integrates time series decomposition (Prophet) with natural language processing (NLP) techniques applied to a six-year corpus of Bolivian news coverage. The methodology employs vector semantic embeddings and zero-shot classification models to capture signals of discursive escalation prior to the materialization of the roadblocks. Using an expanding walk-forward validation scheme applied over 1,762 days and seven forecasting horizons (H+1 to H+7), seven internal configurations and four external benchmarks were compared, including SARIMA and LightGBM. The results demonstrate that the hybrid configuration (Prophet + NLP, C6) consistently outperforms purely statistical models, achieving an AUC-ROC of 0.677 at H+1 and reducing the Brier Score by 10.9% relative to the baseline temporal model (0.220 vs. 0.247), maintaining a statistically significant error reduction across all evaluated horizons ($p < 0.02$). This research validates that the integration of semantic news signals allows for the detection of social tension peaks not captured by historical inertia, providing a technical tool for risk management in critical transport corridors.

Document packets, multiple documents concatenated into a single file, are common in government and administrative workflows, yet splitting them into their constituent documents is difficult, especially for low-resource languages. We introduce Khondo (Bangla for split/segment), the first benchmark for document packet splitting on Bangladeshi government forms. Unlike prior English and OCR-text-based datasets, Khondo is bilingual (Bangla–English) and vision-native; where models operate directly on page images. It spans five concatenation schemes, from sequential to fully shuffled, across 14 administrative domains, with ground-truth boundaries, domain types, and page order. Zero-shot evaluation of MLLMs shows they cluster pages into their source documents fairly well but struggle in restoring the original page order once shuffled. To isolate what drives this difficulty, we run two controlled analyses, varying the prompt instruction and then the packet language. Both primarily affect ordering rather than clustering: (a) explicit page-order instructions are necessary but insufficient, and (b) English packets are ordered more reliably than Bangla, making page arrangement the dominant challenge and language a secondary but consistent factor. Khondo establishes page-order reconstruction as a key open problem in vision-based, low-resource document understanding, and provides a controlled benchmark for measuring progress toward solving it. Our dataset and code is available at https://huggingface.co/datasets/Mausul/khondo

Large language models may infer demographic attributes from subtle linguistic cues even when those attributes are not explicitly stated. This pilot study examines whether Qwen2.5-7B-Instruct internally represents Colombian identity, socioeconomic status, or stereotype-related information when processing Colombian-Spanish and English prompts. We use Natural Language Autoencoders (NLA) to verbalize residual-stream activations from layer 20 across four positional quartiles per prompt. Our dataset contains 30 prompts arranged as 15 matched Spanish-English pairs, spanning explicit Colombian cues, implicit Colombian cues, and neutral controls. We report descriptive rates and qualitative evidence rather than statistically powered effects, focusing on whether latent nationality or stereotype representations appear before they are verbalized in the model output. This work connects activation-level interpretability with bias evaluation for underrepresented Spanish varieties.

Teachers, conference chairs, and public readers all judge writing from limited evidence, seeing only a finished document and not the process that produced it. Final text alone cannot reveal whether a document was produced through human typing, AI generation, or mixed human-AI collaboration. Existing process-tracking tools help, but many are tied to host-document histories, provide coarse activity records, and offer limited control over the writing environment. Humanly is a writing platform that makes the writing process itself the evidence. Users configure writing environments for personal documents or assigned tasks and draft in a workspace that records writing activity and in-platform AI assistance. Humanly can package a completed session into a sealed writing certificate with configuration-aware anomaly behavior review. It can support writing scenarios such as course assignments, peer review, and personal certification. Our user study shows that Humanly is helpful across roles, and a red-teaming study shows that the Humanly Typing Detector distinguishes human hand typing from automated typing.

Pruning a long context means committing to the blocks a model will keep, and a common selector is distilled from a dense teacher's attention. That assumes attention shows which context the answer depends on. We test the assumption on retrieval tasks where the evidence is known exactly, by masking context and measuring whether the answer changes. Attention and causal dependence disagree. Teachers attend to outdated facts the answer does not use, and a selector distilled from that attention routes a multi-hop task at 36% to 98% across runs. The same selector trained on causal evidence sets reaches 99% or better on every run, and known evidence labels beat attention-imitation and accumulated-attention eviction through 32B when the context must be pruned before the question arrives. A reconstruction compressor, KVzip, is stronger on natural-text QA with native per-head KV eviction. Under our paragraph-level adaptation, its scores do no better than labels recovered by intervention.

A systematic review begins with someone reading thousands of abstracts to identify the few that are relevant, and classifiers are used to prioritise that reading. Their inputs are often augmented with Medical Subject Headings (MeSH), assigned either by expert indexers weeks or months after publication or by automatic tools at once. We did not identify prior work comparing the two directly as classifier features, or asking whether that comparison's outcome depends on how the classifier is evaluated. Using the Cohen et al. (2006) drug-class benchmark, we compare expert assignment against one mechanical procedure, substring matching against a MeSH vocabulary drawn from the benchmark, across a bag-of-words logistic regression classifier (seven reruns) and BiomedBERT (five seeds) on three topics. Under the canonical 5-fold full-corpus design the bag-of-words gap on Statins is +0.096 WSS@95%. Stratified subsampling to matched corpus size (n=803) reduces it by roughly two thirds, to +0.033, with a bootstrap interval that includes zero; 10-fold cross-validation at full corpus size reduces it by roughly four fifths, to +0.021. BiomedBERT under canonical evaluation gives +0.020, a difference of 0.001 from the bag-of-words 10-fold result. An empirical power analysis on a single canonical run per topic indicates that a Statins-sized effect at the per-fold variances of the other two topics would not have been detectable at that design (MDE 0.254 for Opioids, 0.384 for ADHD); at the pooled fold count of the multi-run protocol the bound depends on an effective sample size the design does not determine. The results bound the specific lexical matcher tested rather than automatic MeSH indexing in general. More broadly, benchmark conclusions about feature sources can change substantially under reasonable changes to the evaluation design.

Surprisal theory holds that the human processing difficulty of a linguistic unit in context is an affine function of its surprisal under some language model. I argue this claim is a tautology without further constraint: for any non-negative difficulty measure over units in context, there exists a language model whose surprisal is an affine function of it under mild technical conditions. Therefore, because any pattern of difficulty is consistent with some language model, without an additional constraint on the language model, surprisal theory makes no falsifiable predictions. The tautology was long obscured by an assumption implicit in two decades of psycholinguistic work—that the relevant language model is the distribution that generated the training corpus, so that improving corpus fit improves predictions of human behavior. Recent empirical work has undermined this assumption, demonstrating that better corpus models can be worse predictors of processing difficulty. I conclude that breaking the tautology requires a rationalist intervention, i.e., the relevant language model must be derived from a non-empirically motivated model of the comprehender, which could be based on, for instance, memory constraints or processing goals, and that, thus, does not depend on the behavioral data surprisal theory is meant to explain.