Papers

In the context of geodata, existing Large Language Models have often been studied in a homogeneous setting, which has considerably limited insights into their generalization capabilities. In this paper, we present \benchName, a comprehensive benchmark for probing LLMs on geo-related tasks. We leverage a careful selection of twelve publicly available datasets from diverse geo-related tasks and domains, and evaluate a set of LLMs on geo-spatial and temporal understanding using our benchmark. Our results show that reasoning and size have a strong impact on overall performance. GeoBenchLLM is publicly available at https://github.com/Rfr2003/GeoBenchLLM.

Recent agentic reinforcement learning methods use hindsight to complement sparse outcome rewards. However, a completed rollout can yield many such signals, leaving their appropriate allocation across turns unclear. We introduce TRIAL, a trajectory-relative hindsight distillation framework with a unified turn-aligned scoring protocol. For each decision turn, TRIAL extracts an outcome view of that decision's realized consequence and evaluates the same response under ordinary and hindsight-conditioned contexts. The signed log-probability gap determines the direction and local strength of token-level supervision, while turn-level magnitudes are normalized jointly over the realized trajectory. The resulting allocation multipliers have an eligible-token-weighted mean of one, redistributing dense supervision across turns while fixing its average multiplier. Experiments on WebShop and ALFWorld with different backbones show that TRIAL outperforms GRPO across all eight combinations of backbone, environment, and evaluation metric, while achieving the best or tied-best performance among six methods on six of them. On WebShop with Qwen3-1.7B, TRIAL improves the success rate from 56.4% to 75.2% and the task score from 78.7% to 85.7%. Controlled ablations further show that trajectory-relative turn allocation provides substantial gains beyond those of dense hindsight distillation alone.

Scientific literature is increasingly used as a knowledge source for language models, retrieval-augmented generation systems, and research assistants, but answering research questions from papers requires more than fluent generation. A reliable system must identify the relevant papers, locate the concrete evidence that supports the answer, and produce a response that is faithful to that evidence. We present LitTraceQA, a benchmark for literature-grounded question answering over scientific papers. Given a research question and a metadata pool of papers, a system must return three connected outputs: canonical paper identifiers, supporting evidence locations, and answers in one or more requested formats, including free-form text, multiple-choice answers, and structured tables. LitTraceQA targets evidence types common in scientific reading: tables, figures, text spans, equations or algorithms, and citation contexts. The public development split contains 55 examples, including 26 hidden-source single-paper questions and 29 multi-paper questions, and provides gold papers, evidence annotations, and answers for local validation. We also analyze a larger final annotation collection with 4,978 unique-question records over 4,859 unique gold papers. By evaluating paper retrieval, evidence grounding, and answer accuracy separately, LitTraceQA provides a testbed for scientific QA systems that produce verifiable answers rather than unsupported summaries.

Understanding concepts is fundamental to generalization. Despite their impressive performance on a wide range of tasks, Large Language Models (LLMs) still struggle with genuine concept understanding. Prior work has evaluated conceptual understanding in LLMs using natural-language benchmarks or narrowly scoped synthetic tasks, but these settings often conflate multiple skills or lack precise control over the underlying concepts and their properties. To support controlled probing of concepts in LLMs, we design tests on their core properties: abstraction, compositionality, and groundness. We set up a concept-centric benchmark, targeting spatial concepts such as direction, distance, topology, and their compositions, and use question answering tasks serving as a proxy. We conduct extensive experiments across multiple LLM architectures and training regimes to analyze how model scale and design impact conceptual understanding. The results reveal clear limitations in current LLMs and provide insights into the factors shaping their ability to acquire and compose structured concepts. Our findings shed light on how concept-based LLMs can be redesigned for improved information access and knowledge management. The code will be available at https://github.com/rd20karim/concept-probing.

Test data from public benchmarks inevitably leaks into pretraining corpora, inflating evaluation scores once memorized. Contamination mitigation evaluation intervenes in the decoding process to suppress memorization and restore a contaminated model's genuine capability, but its prevailing metric, the G-AP (Gap of Aggregate Performance), is flawed. Discrete correct/incorrect readouts cannot characterize per-question performance, averaging before differencing lets over- and under-suppression cancel out, and uniform per-question weighting invites strategies to push solve probabilities onto the clean model's high-frequency values. We propose SA-PPG (Stratified Aggregate of Per-question Probability Gaps): estimate each question's solve probability by sampling, difference it against the clean model per question, and aggregate within groups defined by the clean model's solve probability. Existing mitigation strategies first estimate where contamination lies and then operate on the estimate, so they are only as correct as the estimate. RailCap instead judges contamination during generation: whenever a sample falls back onto the greedy trajectory, the next trajectory token is capped to the runner-up, accumulating suppression until the response distribution becomes sufficiently dispersed. Across multiple contaminated models and benchmarks, SA-PPG reveals that prior strategies' restoration is substantially overestimated, while RailCap attains the lowest SA-PPG.

Natural Language Processing (NLP) models predicting mental health outcomes rarely specify what they measure: contextual knowledge, emotional content, or syntactic structure. NLP Psychometrics treats psychological prediction from text as a psychometric problem, linking scores to interpretable linguistic evidence and testing beyond the training text format. Nine LLMs, conditioned on controlled personas (cognitive digital shadows), completed psychometric questionnaires with textual explanations per item. We extracted emotional profiles and syntactic-semantic structure via textual forma mentis networks, combined with personality and sociodemographic variables in ablated random forest (RF) regressors, using SHAP to identify which features drove performance and in which direction. Full RF models explained up to 70.8% of variance in life satisfaction (SWLS), 55.7% in depression (PHQ-9), and, for DASS-21, 68.5% depression, 76.0% anxiety, 72.4% stress. Sociodemographics alone explained no meaningful variance in depression, anxiety, or stress, but did so for life satisfaction, where emotion features and income were the strongest predictors; neuroticism and network topology instead dominated depression and anxiety, reversing direction between them. Without retraining, RF models separated diaries from low- and high-score personas ($r$ up to 0.91) and, using only network/emotion features, classified clinical from control participants in real transcripts with up to 68% accuracy. These results show the promise and limits of synthetic data: LLM personas can expose model biases, recover patterns consistent with clinical rumination, and support psychometric prediction from human text without a matched questionnaire, but cannot substitute for human validation. NLP Psychometrics makes these distinctions explicit, measurable, and testable through interpretable AI and network/emotional features.

Grammar engineering requires expertise in linguistic formalism and computational implementation, especially in parallel grammar projects that balance cross-linguistic consistency with language-specific properties. This paper presents the development of Cantonese and Irish treebanks within the Parallel Grammar (ParGram) Project, where linguistic parallelism is maintained at an abstract functional level. We also investigate the methodological potential and limitations of using multilingual LLMs to support grammar engineering, focusing on Cantonese-Irish translation and the generation of formal syntactic structures using OpenAI's gpt-oss-120b model. The results show that translation performance was generally unsatisfactory and unaffected by prompt language. For syntactic structure generation, the model produced some structurally meaningful outputs, but performed poorly on tasks requiring cross-linguistic abstraction. Nonetheless, LLM-generated outputs may still offer some reference value by suggesting alternative analyses and (partially) capturing predicate-argument relations. Overall, our findings highlight both the potential and limitations of using LLMs in collaborative grammar engineering, while underscoring the continued importance of expert-driven analysis and verification.

A central question for computational psycholinguistics is to what extent neural language models can serve as cognitively adequate models of human language processing. However, work has been overwhelmingly focused on the unimodal case of written or spoken language. In contrast, multimodal paradigms, like visual world experiments that present participants with both visual and linguistic input simultaneously, have been neglected. In this paper, we present a novel approach that predicts patterns of gaze preferences in visual world studies. To do so, it derives target preferences by applying an attribution method to a simple bi-modal CLIP bi-encoder. We demonstrate the ability of this approach to qualitatively replicate the gaze patterns found in a seminal English visual world study demonstrating human predictive processing. It does so without a generative architecture and without the need for fine-tuning, despite not being trained for the task.

Large language models (LLMs) can solve complex multi-hop problems yet exhibit puzzling failures on simple two-hop queries: although a model may correctly store each individual hop, it often fails to combine them. To understand the internal mechanisms of this phenomenon, we train transformers from scratch in a controlled symbolic environment. Our experiments reveal a pattern in two-hop generalization: models generalize reliably when the second hop follows the training distribution, but always fail when it deviates. Through mechanistic analysis, we provide a complete explanation for these distinct generalization behaviors: in settings where models generalize successfully, performance is driven by the emergence of consistent intermediate representations for the same entities across contexts, whereas failures on settings where the second hop is out-of-distribution arise from a mismatch across layers: lower layers correctly construct these intermediate representations, but upper layers, while trained on corresponding atomic facts, primarily learn to map them to outputs rather than to reason over them. Driven by this insight, we propose a recurrent-style training strategy, which enables transformers to reuse their reasoning circuitry across input forms and substantially improves generalization on out-of-distribution two-hop queries. Our data and code are available at https://github.com/zzl-strong/two_hop .

Confirmed oncogenic microbes contribute significantly to cancer burden. Identifying novel microbial oncogenicity could yield strategies that will reduce disease burdens. However, relevant evidence is dispersed and infeasible for humans to comprehensively synthesize. LLMs may enable scalable, expert-level systematic evidence synthesis to identify microbe-cancer pairs; however, such capabilities have not yet been demonstrated. Domain experts were recruited to create a dataset to benchmark LLM performance (Gemini 2.5 Pro, Gemini 2.5 Flash, GPT-5, GPT-5 Nano) on 24 research papers using MMTV-LV and breast cancer as a case study. We devised a structured template for evidence extraction and appraisal, consisting of MCQ, Likert-scale, multi-select, and free-text question types (77 items across 24 papers). Agreement between (1) experts and (2) experts and each LLM was determined per question instance using novel metrics. LLMs were assessed by comparing inter-expert and expert-LLM agreement distributions to determine whether LLMs behaved as additional experts by increasing or maintaining inter-expert agreement. Free-text responses were further evaluated qualitatively. Across all question types, LLM responses aligned closely with experts, with GPT-5 and GPT-5 Nano achieving score distributions indistinguishable from experts. Gemini models behaved similarly but were significantly more lenient in applying microbial oncogenesis criteria. Hallucinations were rare. Methodological appraisal and identification of contradictions within full-texts were the most persistent LLM vulnerabilities. GPT-5 and GPT-5 Nano were indistinguishable from experts on structured domain research paper evaluation tasks. This supports use of LLMs for automated systematic evidence synthesis. However, methodological appraisal tasks and contradiction identification in full-texts remain weaknesses requiring strengthening.

We introduce Stoicheia, a 405M-parameter character-level masked-diffusion encoder for Ancient Greek whose input factors into five aligned, independently maskable planes: letters, word and sentence boundaries, diacritics, capitalization, and punctuation. A single backbone can therefore restore lacunae, re-segment, accentuate, and punctuate unspaced text without task-specific retokenization. We pretrain it on an open, revision-pinned corpus of 380M words and release eleven checkpoints: ten rotated, decontaminated folds, guaranteeing that for any given literary passage at least one released model has never seen its text, and one with no exposure to documentary texts. Three experiments - reconstruction of damaged inscriptions and papyri, morphosyntactic tagging and dependency parsing, and macronization with metrical scansion - each carry a matched random-initialization control, isolating what character-level diffusion pretraining contributes: 5.6 CER points on inscription reconstruction, 12.9 LAS on parsing, and 6.0 points of balanced accuracy on macronization. On Ithaca's own test split, with identical frozen samples and strict scoring, Stoicheia reduces character error relative to both prior state-of-the-art systems, from 24.6 (Ithaca) and 23.5 (its 2025 Aeneas-framework successor) to 15.5, and raises top-1 accuracy from 63.0 and 64.0 to 74.5.

Generative models are often evaluated through singular artifacts, whereas human creativity typically emerges through iterative generation, appraisal, and refinement. This pilot study examines whether iterative search improves LLM creativity by adapting FunSearch to recipe generation for the 2024 Pillsbury Bake-Off and evaluating outputs against human benchmarks using TTCT-based LLM evaluation. Across two experiments, we test iteration count, generator temperature, and in-loop selection-scorer model size. Results show that iterative generation-selection can produce recipes with creativity scores comparable to human benchmarks, but additional iterations alone do not improve creativity. The in-loop evaluator matters most: a smaller selection scorer yields significantly higher scores across most TTCT dimensions, while temperature has limited effects except for originality. These findings suggest that evaluator design is a first-order design variable in subjective creative search.

Neural scaling laws are foundational for language model development, yet standard formulations systematically under- and overestimate loss at data-scarce and overtraining extremes. This failure originates in the underlying assumption that model size and training data impact the loss independently. To address this, we introduce the Skaling law, a generalized functional form that couples model capacity and data through a single interaction exponent. This simple extension reduces the Mean Absolute Percentage Error (MAPE) by 1.5-3x across both interpolation and extrapolation regimes. When paired with a sparse grid strategy restricted to low-compute regimes, the Skaling law achieves accurate full-grid extrapolation using approximately 10x less compute than uniform sweeps. By enabling reliable performance prediction from small-scale experiments, the Skaling law provides a more robust and resource-efficient framework for allocating compute budgets in next-generation model training.

Test-time scaling can correct difficult text-to-SQL queries, but the extra computation is normally discarded after each answer. Systems increasingly retain verified repair episodes, yet evaluations still report one end-to-end score. It cannot distinguish replay on recurring questions from help on unseen questions, or identify the responsible memory choice. We call measuring this future value the crystallization problem. Our controlled evaluation holds the single-shot solver fixed and varies one memory choice at a time. We separately measure replay, cross-question retention, and held-out same-database transfer. On BIRD, storing verified corrected queries improves held-out first-attempt accuracy by 4.34 percentage points. This gain captures 44.4% of the accuracy headroom provided by on-demand repair on the same questions. Controlled interventions identify database-specific content as the main operating ingredient. Reliable verification and broader retrieval coverage yield supported gains; richer formats and elaborate retrievers do not. Open-source code, evaluation artifacts, and reproduction instructions are available at https://github.com/ai-jiaqian/text-to-sql-memory-crystallization.

Recent work shows that Large Language Models (LLMs) encode more information in their activations than they show in their responses. We ask whether the information contained in activations could be used as a source of measurement for the presence of concepts. We compare two algorithms for extraction of concepts, i.e., Recursive Feature Machine (RFM) algorithm and linear probing against an embedding baseline, surface baselines, and the model's own response using a human-annotated Environmental, Social and Governance (ESG) dataset of financial texts. The best linear probe ranges from 0.6 to 3.3 percentage points below fine-tuned domain classifiers. The probes also outscore the model's own response in all comparisons. Furthermore, the probes outperform the RFM concept vectors in all accuracy comparisons. In a synthetic mixture test, their mean-pooled score tracks the number of concept sentences in a paragraph better than the concept-vector score. Put together, a linear probe on a frozen LLM is a cost-effective and generalizable alternative to fine-tuning a domain classifier.

Scientific claim verification over a cited paper requires predicting the claim–paper relation and identifying the paragraphs that justify that prediction. This setting poses two linked challenges: within-paper distractors often resemble genuine evidence, while a classifier trained on gold evidence must operate on retrieved evidence at inference. We present HNR-DAC, a two-stage framework that trains each stage on the cases it will actually encounter. Hard-Negative Reranking (HNR) quantifies evidence confusability using a base reranker's scores on non-gold paragraphs and contrasts gold evidence against the most confusable candidates. Distribution-Aligned Classification (DAC) trains on the Top-1 paragraph produced by the same frozen HNR used to construct inference inputs, while HNR's Top-3 paragraph identifiers provide the evidence output. On the NLPCC 2026 Task 10 Track 2, the final configuration obtains 97.21% Hit@3, 95.79% Macro-F1, 94.47% Joint@3, and an average score of 95.13%. The corresponding submission ranks third on the official Track 2 leaderboard while achieving the highest overall Macro-F1 of 93.05%, alongside 70.16% Joint@3 and an average score of 81.61%.

Test-time training (TTT) views sequence modeling as an online learning problem in which fast weights are updated by an internal learning rule. Despite the growing number of TTT variants, existing approaches typically hard-code each variant separately, which makes it difficult to design new TTT methods and to isolate the role of each component. To address this, we propose Modular TTT, a framework that represents the inner learner as a directed acyclic graph and exposes the fast-weight network, loss function, learning rate, weight decay, and normalization as explicit design dimensions. Modular TTT automatically composes primitive-level train-view forward, train-view backward, and causal query-view rules into the full graph-level TTT computation, including the fast-weight state transition. Using Modular TTT, we systematically ablate the components of TTT and find that small learning-rate initialization, weight decay, and a single-layer nonlinearity improve performance, while MSE and inner-product losses perform similarly. Deeper fast-weight networks and normalization tend to hurt performance because they induce excessively large activations, while residual connections and gating provide little measurable benefit. Guided by these findings, we train the best resulting variant as 410M- and 1.45B-parameter models on 100B tokens, and observe training loss and benchmark performance comparable to Gated DeltaNet.

Long-document understanding requires locating sparse and heterogeneous evidence across hundreds of pages, yet existing systems remain limited by static retrieval and fragile cross-round memory. Mainstream single-round methods commit to a fixed top-$k$ page set at the outset and struggle to recover from early retrieval errors; recent iterative approaches allow multi-round evidence acquisition, but they do not investigate the propagation mechanism of cross-round states, making it difficult to track the dynamic changes in page relevance. To address these limitations, we propose DocMemo, a memory-guided framework that formulates long-document reasoning as dynamic evidence exploration. DocMemo maintains a tri-level retrieval state consisting of Document Schema Memory, Page Belief Memory, and Question Episodic Memory, which respectively capture structural priors, dynamic relevance estimation, and query-specific reasoning trajectories. During reasoning, DocMemo continuously refines cross-round page selection through Bayesian page belief updating with Thompson sampling, spatial proximity propagation, and structure-aware adaptive-granularity evidence access, while supplementing page-level evidence with fine-grained visual regions. Experiments on 3 benchmarks show that DocMemo achieves state-of-the-art performance and validate the efficacy of structured memory and dynamic page belief updating. Code is available at https://github.com/Harrygof/DocMemo.

Organizing thousands of unstandardized, multilingual expertise declarations is a persistent challenge for Human Resources (HR) platforms, directly impacting downstream tasks like accurate talent matching. To address this, we propose a hybrid knowledge graph generation pipeline that grounds a Large Language Model (LLM) in the Wikidata multilingual Knowledge Graph (KG) while employing an agentic reflexion pattern to synthesize emerging concepts and their associated metadata. Unlike rigid top-down methods or fragmented bottom-up approaches, our system anchors recognized concepts to stable Knowledge Graph entities while dynamically creating new nodes and relational metadata for unrecognized skills. Executed across five stages, entity reconciliation, multilingual canonicalization, active curation, deduplication, and the iterative recovery of unmapped concepts, the system autonomously adapts to rapidly evolving, noisy skill mentions across five European languages. Ultimately, this pipeline provides a highly scalable, explicable, and self-healing framework for generating a comprehensive skills knowledge graph, from which a structured taxonomy is derived, using unstructured, noisy text.

Visual retrieval-augmented generation (RAG) commonly expands the retrieved evidence set to improve answer-page coverage, implicitly assuming that all available evidence should be passed to the generator. We show that this assumption does not hold for diffusion language models (DLMs): retrieving more pages increases answer-page recall, whereas unconditionally passing all retrieved pages to the generator often reduces answer accuracy, primarily because of semantic conflict. A latent-source analysis explains this mismatch through source-coherence loss in parallel denoising, where position-wise proposals can combine incompatible visual sources into unsupported answers. We further find that such interference is already visible in the first-step answer-block distribution, making it possible to assess evidence before decoding. To preserve retrieval coverage while limiting harmful visual exposure, we propose the Entropy-Based Candidate Filter (ECF), a training-free evidence-admission framework. To reduce irrelevant content within individual candidates, ECF constructs multi-granularity evidence units; to identify beneficial additional evidence, it uses blank-controlled block confidence and retrieval rank to determine whether and which candidate should enter the final context. Across three multimodal DLMs and five visual QA benchmarks, ECF improves answer accuracy by 2.62 percentage points on average over the strongest fixed top-$k$ input and, with LLaDA2.0-Uni, by 2.37 percentage points on average over the best competing training-free result for each dataset. These results show that broader retrieval benefits visual DLM-RAG through selective evidence admission rather than unconditional evidence expansion. Code is publicly available at https://github.com/wjkuser/ECF.

We present a web demo for exploring a large-scale disambiguated knowledge base (KB) materialized from a large language model (LLM). GPTKB 2.0 contains 38.4M triples over 1.6M canonical entities, together with 207.6K consolidated relations and 66K consolidated classes. Unlike prior LLM-derived knowledge bases that largely identify entities by surface strings, GPTKB 2.0 performs context-guided disambiguation during recursive KB construction, separating homonyms and merging synonymous mentions as facts are elicited. The demo makes this process inspectable: users can browse entities, follow links across the KB, and audit the provenance of individual facts, including surface forms, candidate matches, source triples, and disambiguation decisions. The interface further supports structured SPARQL queries, natural-language questions translated to SPARQL, and entity linking from user-provided text to canonical GPTKB 2.0 entries. GPTKB 2.0 is available at https://gptkb.org/, with the full KB downloadable for offline use.

It is well established that large language models (LLMs) are sensitive to prompt framing, reflecting patterns in their training data or prior prompts. In this study, we investigate the extent to which LLMs reinforce users biases expressed in the prompts and examine the boundary between implicit framing effects and explicit prompt manipulation. Specifically, we evaluate how susceptible LLMs are to direct and suggestive prompts that encourage models to support or challenge particular positions. We evaluate six LLMs using 160 distinct prompts spanning ten topics across opinion-based and factual domains. The prompts systematically vary in prompting strategy, support versus challenge instructions, prompt polarity, users' expressed beliefs, and topic domain, spanning both opinion-based and factual questions. Our results show that LLMs systematically adapt their responses to align with prompt framing, even in factual contexts. This suggests that prompt framing can outweigh factual consistency in model responses. Overall, our findings delineate the extent and boundaries of LLM manipulability. Furthermore, the results imply that LLMs can reinforce subtle user biases and are susceptible to explicit prompt manipulation even in domains where responses should remain factually stable.

Long-horizon role-playing demands that characters remain recognizable as they evolve with the narrative. Yet existing work falls short on two fronts: representations are typically static profiles that cannot be updated locally without destabilizing unchanged traits, and benchmarks mainly test persona preservation and memory recall rather than whether a model speaks from a character's currently evolved state. We address both. PHASE-Tree is a multi-timescale character-state tree with an immutable identity root and mutable persona, session, and moment layers, making each mutable field an addressable target for localized within- and cross-episode updates. It conditions generation through explicit textual provision or implicit parametric adaptation. To measure evolved-state generation, we introduce LongEvoRoleBench, which pairs four long-dialogue corpora for cross-episode evolution with four short-dialogue corpora as within-scene state-tracking checks, under a unified next-utterance protocol. On the long-dialogue core, textual PHASE-Tree ranks first in 11 of 12 dataset-metric cells against internal variants and all 12 cells against external textual baselines, improving character-level, semantic, and embedding scores by 19.7%, 12.4%, and 15.1% respectively. In a blinded 200-response study, human ratings correlate with the GPT-4.1 judge (Pearson r= 0.65); on descriptive n= 10 PT and NR prompt subsets, the Overall difference is +0.20. The long-dialogue Sem advantage persists across LLM judges and generation backbones.

Current evaluations do not isolate whether text-only language models can originate visual concepts before image generation. Fluent visual prose can hide visual-plan failures: an answer may appear creative while repeating familiar visual clichés or failing to specify a renderable scene. We define Visual Creative Ideation (VCI) as the ability to produce textual visual plans that are useful, expressive, and population-novel, and introduce Ekphrasis, a 400-task benchmark spanning Abstraction, Combination, Transformation, and Adaptation. Ekphrasis scores anonymized pairwise comparisons with dimension-specific checklists, aggregates preferences with Bradley-Terry models, and uses Typed Idea Graphs to convert task-specific population clichés into novelty references. Across 14 language models, VCI separates usefulness, expressiveness, and novelty rather than reducing to fluency: strong models achieve similar overall scores through different profiles, and useful plans can remain visually clichéd. A cross-modal grounding study further shows that text-level VCI ordering largely survives faithful rendering and blind image-level preference judgment, supporting Ekphrasis as a measure of visual ideation beyond prose quality.

Agent memory systems compress what they store, and compression is built to drop qualifiers, so a claim's epistemic standing tends not to survive being written to memory. We ask what governs whether it does. Matched notes carry the identical claim and identical stance and differ only in where that stance sits; one model compresses both under the same budget among the same filler notes, and a blind reader that never sees the condition scores the result. Across 60 claims in seven registers, writing the stance as a labelled field rather than a bracketed aside raises retention by about 15 points on two models (37 claims to 2 on one, 30 to 8 on the other; permutation p=0.00005), and a pre-registered replication on Haiku, its prediction and decision rule committed before the run, gives +15.6 points, 38 claims to 1. Ablating the format on both models gives the same net effect from different parts: labels help on both (+9.7 and +12.8) and length helps on neither, but wording the stance as a full sentence is the largest component on one model (+12.5) and worth nothing on the other (+0.6). Either model alone would have licensed a confident and different mechanism, so we claim only the intersection: make the stance explicit, not merely longer, and expect the best way of being explicit to depend on the model. A deterministic readout with no model reproduces the two-cell direction and five of seven ablation contrasts, but not length or labels, which we therefore do not claim on one instrument. Fifty hand labels (kappa=0.75) agree on direction; we print their seven disagreements in full. We also report nine withdrawn claims, three of them former title claims of this paper.

Today, we improve models by training and evaluating them on problems at the frontier of their abilities. Creating such problems is itself a demanding task, requiring the ability to probe model limits and generalize beyond existing question distributions. It also means placing problems at a precise difficulty level, which requires understanding what it takes to solve them. In short, generating problems calibrated to a model's current frontier demands capability beyond it, an increasingly burdensome constraint as models improve. Our key insight is that we can leverage this constraint to our advantage: a model that can generate problems consistently calibrated to a given frontier must possess capability beyond it. Accordingly, we present Ask-E, an environment that benchmarks and trains models on their ability to write questions at a given skill level, rather than answer them. Concretely, we define target skill levels as ranges bounded by the capabilities of two existing language models. A generated question is successfully calibrated if exactly one of the two models can solve it, placing it precisely within the target range and differentiating the capabilities of these models. Ask-E serves both as a benchmark and a training environment, where models generate problems calibrated to a variety of skill levels. We find that even frontier models achieve below 50% calibration on the benchmark, leaving significant headroom to measure future progress. We also show that training on this environment leads to improvements across a number of downstream math benchmarks even with no new math data, no interaction with stronger models, and no correctness-based reward.

Designing autonomous agents that effectively assist human teams hinges on understanding team dynamics, often without task specific knowledge. We present TRIBE, a domain independent approach that reveals team behavioral dynamics invisible to traditional performance metrics. We show that communication patterns can categorize teams into performance predictive behavioral tribes, as early as 10% into the task, enabling timely interventions. We test TRIBE on four diverse datasets and demonstrate that communication patterns predict team performance while the prediction strength varies by the degree a task structure allows for behavioral freedom. Our temporal analysis reveals that AI agents significantly alter team behavioral trajectories while human advisors align with natural dynamics, and that teams maintain behavioral flexibility throughout collaboration. Further, we compare TRIBE to Llama and optimize the pipeline, achieving significant speedup with performance improvement.

We propose Zero-phase Component Analysis (ZCA) whitening as a geometric pre-processing step for the Word Embedding Association Test (WEAT). WEAT is a bias measurement method widely used in both computational social science and AI fairness research. It relies on cosine similarity as a measure of semantic association, which assumes that the embedding space is approximately isotropic. However, prior work has reported that many widely used language models do not satisfy this assumption, raising concerns about the reliability of bias measurements. ZCA whitening transforms the covariance of the embedding space into the identity matrix while minimizing perturbation to the original vectors. This transformation restores the isotropy condition on which WEAT relies. We evaluate our approach on ten standard WEAT test suites and seven models spanning three architectural families, yielding 70 model-task combinations. The results show that ZCA whitening substantially reduces the anisotropy of the embedding spaces across all models. Particularly for highly anisotropic models, we further observe improvements on standard semantic similarity benchmarks, indicating that the calibrated space better captures semantic associations. After calibration, over 30% of WEAT results change significance status, and effect sizes shift in both directions depending on bias category. These shifts suggest that uncalibrated measurements may both overestimate and underestimate the associations encoded in the embedding space. These findings indicate that previously reported bias measurements in anisotropic embedding spaces should be interpreted with caution and may benefit from re-evaluation with calibrated methods. Our approach contributes to restoring the measurement foundation of WEAT across both computational social science and AI fairness research.

Researchers who seek to build social robot applications on foundation models are faced with a difficult question: how should we pick a model? Public leaderboards offer little guidance: the demands of real-time, embodied social interaction lie largely outside their focus. And direct evaluation is impractical at scale: each embodied study requires scarce participant, robot, and experimenter time. In this paper, we identify five evaluation dimensions for foundation models in social robots: (i) conversational competence, (ii) user safety, (iii) embodied character, (iv) target scene effectiveness, and (v) audience appropriateness. To make model selection cheaper and better informed, we propose a three-tiered evaluation funnel paradigm that first filters with general metrics, then extends to simulated interactions, and terminates in more expensive, robot-specific evaluation. We map all five dimensions across all three tiers, chart where applicable evaluation methods exist and are missing, and close with a call to action: let's build the evaluation framework together as a community.

Biological specimen records collected by natural history institutions constitute a rich source of temporal geographic knowledge, capturing biodiversity information about regional landscapes as they were recorded at different times. Using digitised data from the Allan Herbarium (New Zealand), this study identifies place names in these specimen locality descriptions that are absent from current gazetteers; we refer to these as non-gazetteer place names (NGPs). These place names are typically historical, vernacular, or colloquial and were used as landmarks to describe a specimen's location at the time of collection. We then investigate the problem of georeferencing the NGPs using only the limited information available in the specimen records. To resolve this, we leverage repeated occurrences of the same place name across specimen records with different specimen locations and spatial relation terms, extracting and inverting these relations to derive constraints on NGP locations. This approach is instantiated within deterministic, probabilistic, and LLM-based methods, enabling a comparative analysis of their strengths and limitations for text-based spatial inference. On a pseudo-NGP benchmark, probabilistic inference achieves the highest accuracy (median error 1.43 km; A@1 km 36%), while the LLM yields competitive but less precise estimates (median error 1.80 km; A@1 km 31%), indicating that, despite advances in LLMs, traditional modelling remains advantageous when high spatial precision is required.