QA
-
Sweep Less, Steer More: Using Activation Steering Geometry for Reduced Hyperparameter Search
2026-11-01 03:29:29Nishkal Hundia, Swastik Agrawal, Navita Goyal, Sarah WiegreffeAbstract
Representation steering methods like difference-in-means (DIM) offer a lightweight approach to controlling LLM behavior, but their effectiveness varies widely across concepts, datasets, and layers, making principled deployment costly without exhaustive hyperparameter sweeps. We build on the discriminability index $d'$, a compute-cheap geometric measure of how well contrastive training activations separate along the steering direction. Across five behaviors on 3 models, we demonstrate uses beyond its established role as a dataset-level steerability predictor. First, in the multiple-choice question answering (MCQA) setting, we show that $d'$ reliably identifies performant layers at which to steer, eliminating the need for a layer-wise sweep. Second, we extend this analysis to open-ended multi-token generation, and find that $d'$ remains successful for layer selection across different settings. Lastly, we test whether the position of a generated token's representation on the DIM line predicts steering success on that individual example. In the MCQA setting, this works for high-$d'$ datasets but is inconsistent across behaviors, models, and training-token choices in open-ended generation. This highlights that methods developed and conclusions drawn in single-token MCQA settings do not necessarily transfer to open-ended generation.
Abstract (translated)
URL
https://arxiv.org/abs/2611.01496
PDF
-
Can Large Language Model Agents Learn by Watching Tutorial Videos?
2026-11-01 02:48:09Haoyu Zhao, Zafeirios Fountas, Jun WangAbstract
Humans continuously acquire knowledge from carefully crafted materials such as books and tutorial videos, carrying that understanding into future tasks. Millions of such instructional videos are uploaded online every day, yet LLM agents remain unable to exploit this ever-expanding wealth of human teaching. Their capabilities are frozen after training, while parameter updates at scale remain prohibitively expensive. This raises the central question: can LLM agents learn from this vast instructional resource without parameter updates? We introduce Tutorial Agentic Learning (TAL), a nonparametric continual-learning paradigm that consolidates tutorial videos into a persistent, structured tutorialbook memory through a two-stage adversarial process that harvests hard, video-grounded questions and uses them to refine the memory. At inference time, TAL retrieves through a Table-of-Contents (TOC) tree, tying retrieval directly to the model's reasoning rather than to embedding similarity. To study this setting, we introduce VidLearn, a multimodal QA benchmark built from 72 chess tutorial videos spanning 12 openings. On VidLearn, TAL outperforms all evaluated baselines and improves monotonically as more videos are consumed, achieving up to 113% and 91% relative gains over the corresponding vanilla MLLMs on the training and test splits, respectively. In actual gameplay, TAL reaches a stronger opening-phase position than its vanilla counterpart in 69% of head-to-head games. Evaluation in a second domain further demonstrates the generality of TAL. Together, these results highlight instructional videos as a scalable resource for parameter-update-free agent continual learning.
Abstract (translated)
URL
https://arxiv.org/abs/2611.01460
PDF
-
Wrong Together: Shared Bias in User Simulation
Abstract
User simulators generate responses in place of human participants, allowing researchers to estimate the result of a study. To use these estimates before collecting human responses, researchers need a way to assess their reliability. We test whether disagreement among simulators can identify unreliable estimates of a group's response proportions and of differences between groups. We evaluate this signal using 6 language-model simulators and existing GlobalOpinionQA responses for 60 questions in 12 countries. Disagreement predicts invalidity more accurately for differences between two groups than for a single group's response proportions (AUC 0.87 versus 0.71). Errors shared by the simulators remain invisible to this signal. When such an error affects both groups equally, it cancels in the comparison, helping explain the stronger result for group differences. We test this explanation by estimating the shared bias using one set of groups and subtracting it from predictions for a separate set. This correction reduces prediction error without changing disagreement among the simulators. In a retrospective analysis, we simulate referring the 20% of group comparisons with the greatest disagreement for human assessment. Among the remaining comparisons with a conclusive validity judgment, 54.8% are judged invalid, compared with 63.7% before referral.
Abstract (translated)
URL
https://arxiv.org/abs/2611.01354
PDF
-
Geometry of Confabulation and Refusal in Small Language Models
2026-10-31 21:14:26Vladimir KazeiAbstract
Faced with an unanswerable question, small language models can either refuse or confabulate: confidently produce a wrong answer. Can the hidden state before the first generated token provide a clue? Can we flip confabulations into refusals? I introduce ConfabQA, a 784-item benchmark whose obscure-pre-cutoff category breaks the typical correctness/knowledge-cutoff confound and map the geometry of the hidden state with linear probes, a sparse autoencoder (SAE), and direct optimization of intervention directions. On Qwen3-1.7B and Gemma 2 2B the probe separates upcoming refusals from confabulations well beyond prompt features (+7.4/+16.3 pp); on Llama 3.2 3B the ConfabQA margin is within fold-to-fold spread, but a direct refusal probe on PopQA and TriviaQA recovers +18-25 pp. On Qwen I closely inspect three candidate directions for the refusal vector: probe-recovered, Qwen-Scope SAE feature, and directly optimized. Although the three directions are pairwise 50-81 degrees apart, a push at or below the state's own norm along any of them flips most confabulations into refusal openings dominated by the "As..." family. Forcing the opener without touching the hidden state yields the same one-in-three judged-refusal rate as any push. The opposite push collapses refusals mostly into confabulations, yet leaves originally correct answers intact, so the direction gates the opening speech act rather than the knowledge behind it. Confabulations do not resolve into a concentrated feature in a dictionary or a token signature on Qwen3-1.7B. Scoring every probe on the margin over the strongest of four prompt-feature baselines extracts the hidden state's own contribution. Correctness signal is model-dependent and portable only for Llama 3.2 3B: 81% of the probe's margin over test majority survives transfer between ConfabQA, PopQA, and TriviaQA. Benchmark, judge, and code are released.
Abstract (translated)
URL
https://arxiv.org/abs/2611.01272
PDF
-
MedAstra: ASR-less Speech Retrieval with Diagnostic Grounding for Medical QA
2026-10-31 21:00:22Elahe Khatibi, Ziyu Wang, Amir M. RahmaniAbstract
Spoken medical question answering can fail when automatic speech recognition (ASR) corrupts a clinically important term before retrieval. We present MedAstra, a framework that maps frozen WavLM speech representations through a lightweight adapter into a frozen E5 text embedding space. This enables a dense retrieval path that accepts a waveform without a query transcript. Retrieved passages and speech-linked textual metadata support answer generation with entity alternatives, diagnostic contrasts, relation heuristics, and bounded retrieval refinement. ASR-less refers to the initial speech retrieval interface; lexical retrieval, string-based entity correction, and the text generator require textual inputs. The deterministic speech embedding does not constitute a calibrated model of acoustic uncertainty. Reported retrieval results reach 87.6 Recall@10 for the speech-only retriever and 90.4 for the hybrid configuration, compared with 81.0 for transcript-first dense retrieval. Medical QA and ablations provide descriptive system comparisons, with input-provenance and evaluation limitations stated explicitly. We distinguish controlled synthetic acoustic tests from real clinical speech, qualify component attribution, and specify the remaining comparisons needed to assess speech-only QA and clinical generalization.
Abstract (translated)
URL
https://arxiv.org/abs/2611.01261
PDF
-
CompSkill: Compiling Execution Experience for Skill Retrieval
2026-10-31 17:00:57Qi Shen, Ziyue Zhu, Yongqing Jiang, Zhen Xia, Li Shengjie, Shuai Zhao, Bo Cheng, Xiaobao Wu, Anh Tuan Luu, Haoran LuoAbstract
LLM agents increasingly rely on skill documents, yet most retrievers rank them by description alone and ignore what earlier executions revealed, even though a relevant skill can hurt execution. We present CompSkill, which indexes skills with learned 512-d keys and compiles matched same-task comparisons into skill offsets and capped key residuals (KeyOpt), so experience changes which document the executor receives, never its text. Under a logistic outcome model with unknown task difficulty, matched comparisons cancel that difficulty and only discordant ones carry information; on support executions, predictions improve with each discordant task. An experience hypergraph keeps each comparison with its executor and document versions, recalls missed skills and grounds a cited review. Compact keys raise SkillRet NDCG@10 from 64.0 to 70.9 at one eighth of the dimension. With one document, CompSkill completes 256 and 260 of 332 ToolQA tasks with GPT-5.6-luna and DeepSeek-V4.1-Flash, respectively, more than every baseline, and stays on par with the strongest baselines on AppWorld and ALFWorld. On 332 newly generated ToolQA tasks, it completes 256 with GPT-5.6-luna, significantly more than BM25, SkillRet-8B, Tool2Vec and RankGPT; over all 844 ToolQA tasks, it completes 24 more than the strongest baseline. Selection matters beyond relevance: pairing the right guide with a same-database sibling instead of another database's guide costs 8 of 96 tasks. Code, models and support data: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2611.01115
PDF
-
Labadain-MedQA: A Tetun Medical QA Dataset for Fine-Tuning and Evaluating Large Language Models
2026-10-31 13:36:05Gabriel de Jesus, S\'ergio NunesAbstract
Advances in large language models (LLMs) have created opportunities for cross-lingual transfer of medical knowledge from high-resource to low-resource languages to support access to healthcare information. However, medical question-answering (QA) datasets for fine-tuning and evaluating LLMs remain scarce for Tetun. To address this gap, we translated MedAlpaca's Medical Flashcards and WikiDoc Patient Information collections into Tetun using Gemini 2.5 Flash, producing 39,077 medical QA pairs. Translation quality was assessed on 50 randomly sampled QA pairs (100 documents) using automatic metrics against human reference translations and evaluations by two native Tetun-speaking healthcare professionals, focusing on medical accuracy and Tetun fluency. Human translations achieved higher mean ratings than LLM-generated translations for both criteria (4.43/5 vs. 4.02/5), with statistically significant differences (Wilcoxon signed-rank test, p < 0.001). To assess downstream utility, we fine-tuned Gemma and Llama variants on the translated MedAlpaca dataset using parameter-efficient fine-tuning (PEFT) and evaluated them on 325 Tetun-translated United States Medical Licensing Examination multiple-choice questions. Gemma 4 12B fine-tuned using QLoRA achieved relative improvements in mean accuracy of 10.79% with 4-bit NormalFloat (NF4) inference and 10.42% after merging the LoRA adapters into the BF16 base model, compared with the corresponding base models. These results highlight the viability of automatic dataset translation for Tetun medical QA and the effectiveness of PEFT in enhancing model performance. The datasets, models, and evaluation resources are accessible at this https URL to support reproducible research on Tetun medical QA and facilitate machine translation, cross lingual, and multilingual research.
Abstract (translated)
URL
https://arxiv.org/abs/2611.00935
PDF
-
When 200 OK Is Not an Answer: Silent Fetch Failures in Agentic Web Content Acquisition
2026-10-31 12:42:45Shuyang Zhang (The Hong Kong Polytechnic University)Abstract
Web agents and answer engines fetch pages at question time, and liveness checks of cited sources treat an HTTP 200 response as success. We measure how often that assumption fails and what it does to measurements built on fetched pages. We fetched 10,307 URLs in three days with a polite static client and a rendered browser client: pages from 1,507 sites in five strata, random non-existent paths on every site, and the 4,795 URLs cited by expert-verified claims in ExpertQA. We define a taxonomy of what a 200 response contains, label the static 200 responses automatically, and check the labels against 300 items reviewed by an LLM-based assistant; a human check of 60 of them retained every reviewed usability label after discussion. By the automatic labels, 20.9% of the pages that a static client receives with status 200 do not contain the page's main content: app shells, bot challenges, block pages, error and placeholder pages, and partial content. On the held-out test split, the reviewed sample lowers this estimate from 25.5% to 20.0%. 13.1% of the sites whose random-path probe was answered returned 200. The error propagates: for the cited sources, treating 200 as content overstates the share of claim-source pairs that can be checked against live content by 12.2 points and inflates the apparent silent citation-drift rate from 5.8% to 21.9%. Cheap heuristics catch most of it. Text length, probe comparison and redirect-to-root, combined and tuned on held-out sites, recover 86% of unusable reviewed responses at a false-alarm rate near 5%. Content-reading detectors perform comparably: the best, the text classifier, gains five to seven points of recall at more than twice the false-alarm rate. We argue that agents and their evaluations should report fetch outcomes explicitly instead of inferring them from the status line and will release the labelled data and code.
Abstract (translated)
URL
https://arxiv.org/abs/2611.00884
PDF
-
Adaptive-Policy Distillation: Knowledge Injection under Information Asymmetry
2026-10-31 11:59:35Hojun Cho, Kang Eun Jeon, Jaegul ChooAbstract
On-policy distillation (OPD), which uses reverse-KL supervision on student rollouts to transfer a context-conditioned model's behavior into student model parameters for closed-book inference, has shown strong results. However, we find that this success does not translate to knowledge injection: across two document-QA benchmarks, OPD injects document knowledge far less efficiently than simple off-policy distillation. We trace this reversal to two failures that are structural, yet sparse across token positions. Reverse KL weights its gradient by the student distribution and therefore barely lifts document-specific tokens the student considers unlikely; after forward KL restores this signal, student rollouts expose later positions on the target trajectory only after generating the preceding tokens correctly, so coverage deteriorates with target length. Motivated by the concentration of teacher-student divergence at these hard positions, we propose Adaptive-Policy Distillation (APD), whose per-token gate jointly shifts the rollout toward the teacher and the objective toward forward KL where needed, withdrawing both interventions as the student learns. Without labels or a verifier, APD improves over OPD by up to 54.6%, achieving the best scores of 0.558 F1 on QASPER and 0.486 MRR on BioASQ.
Abstract (translated)
URL
https://arxiv.org/abs/2611.00852
PDF
-
From Small to Large: Assessing Medical Reasoning, Benchmark Contamination, and Model Size in Language Models
2026-10-30 23:23:42Navapat Nananukul, Mayank KejriwalAbstract
Large language models (LLMs) now answer medical exam questions at or above expert level, but it is unclear whether such scores reflect clinical competence, or are instead affected by the format of the questions being asked (e.g., multiple-choice versus open-ended), prior exposure to the benchmarks, and the difficulty of the questions. These benchmarks are also public and widely used online, which may be resulting in LLM memorization. Models may perform better on questions that they have seen before. It is also not evident whether smaller models, which are more locally deployable in HIPAA-compliant hospital settings, have achieved human performance. In this study, we evaluate fourteen current models from six leading providers, one smaller and one larger model from each provider with two additional open-weight models, on 3,076 questions from four medical benchmarks (MedQA-USMLE, MedMCQA, PubMedQA, and MedBullets) under standard multiple choice, the same questions with the answer choices withheld, and reconstruction of a hidden question from its choices alone. In the standard multiple-choice setting, the strongest models exceed the human-expert baseline on MedQA-USMLE and PubMedQA, but the smaller models trail their larger models by 8.7 percentage points and only two of the six pass the human-expert bar on the licensing examination. Withholding the answer choices, however, makes every model struggle: mean accuracy falls from 82.3% to 50.9%, and the two largest drops belong to the medically specialised models. On memorization, the strongest models reconstruct one question in four from the answer choices alone, yet accuracy holds on the questions they cannot reconstruct. We further weigh accuracy against inference cost and test whether ensembling several models improves medical question answering.
Abstract (translated)
URL
https://arxiv.org/abs/2611.00460
PDF
-
LF-VideoQA: Evaluating Long-Video Understanding with Long-Form Theory-of-Mind Questions
2026-10-30 21:05:36Shaden Shaar, Bradon Thymes, Majd Aldaye, Claire Cardie, Bharath HariharanAbstract
The hardest part of understanding a long video is not pinpointing any one moment but synthesizing many, e.g., answering a question that requires tracking what develops across hours of content and articulating how the pieces fit together. To date, most long-video QA datasets require only a short answer or a multiple-choice selection, oversimplifying the long-video understanding task to retrieval. We introduce LF-VideoQA, a dataset for long-form, open-ended question answering over long narrative videos built on theory-of-mind questions: a character's belief depends on everything they have seen or been told, so each question demands assembling and reasoning from evidence distributed across the video. The dataset contains more than 6,300 questions on 94 recent films; each requires reasoning over scenes spanning an average of 23 minutes and comes with a question-specific rubric for recall and precision scoring. The dataset is created entirely automatically by annotating screenplays with scene events and each character's first-order beliefs and meta-beliefs, then generating (question, answer, rubric) triples through LLM-based question generation and verification; human studies validate their quality. We evaluate four families of state-of-the-art open-source VLMs across three settings and find the task highly challenging: (i) even an oracle with access to the screenplay falls well short of complete answers; (ii) the best non-oracle model attains barely half of the recall level of the oracle; (iii) unlike on prior long-video benchmarks, captions and subtitles underperform the video itself, indicating that the questions need visual evidence; and (iv) recall is the predominant bottleneck, with answers omitting many required rubric claims.
Abstract (translated)
URL
https://arxiv.org/abs/2611.00383
PDF
-
SPARK: Self-Correcting Embodied Memory via Physical Re-Observation
2026-10-30 20:24:55Dmytro Kurdydyk, Aadi Palnitkar, Sebastian Scherer, Seungchan KimAbstract
Retrieval-augmented generation (RAG) systems have seen growing adoption in embodied settings as language-queryable memories of the environment. However, existing approaches often assume a reliable single-episode knowledge base (KB) or accumulate observations across episodes without correcting prior errors. Such errors can persist through incorrect semantic labels, inaccurate object positions, and inconsistent spatial relationships. To address this, we present SPARK, a corrective multi-episode framework for embodied RAG. After the first episode, a semantic critic flags regions of the KB, a targeted planner generates revisits, and a selective merge adds, removes, or replaces KB information based on the new observations. We evaluate SPARK on four outdoor AirSim maps under two degraded viewpoint conditions, using perception-built KBs, a visibility-constrained oracle used only for evaluation, and a fixed set of 158 embodied question answering (EQA) questions scored before and after correction. A single corrective episode improves downstream EQA on every map under both conditions while storing 13-41% fewer objects than append-only updating, with gains up to 2.7x on one map-condition pair. Extending correction through Episode 4 brings every map above the EQA performance of its undegraded reference. Ablations show that the critic's three signals each contribute to performance, that targeted revisits achieve a larger mean gain than untargeted ones under the same budget, and that selective merging achieves gains comparable to append-only updating with a smaller KB. Overall, SPARK demonstrates the value of correcting embodied memory through targeted re-observation, particularly as embodied systems scale and the risk of accumulated errors increases.
Abstract (translated)
URL
https://arxiv.org/abs/2611.00344
PDF
-
Think in 360{deg}: Benchmarking and Bootstrapping Panoramic Video Reasoning
2026-10-30 17:35:02Lidong Lu, Lidong Zou, Xinyi Mao, Jiaxin Lv, Zhenquan Li, Liangrui Dong, Tong Lu, Guo ChenAbstract
A panoramic video captures all directions around the camera, but relevant evidence can appear in different directions and at different times. Understanding it requires models to connect observations across the viewing sphere and time. We introduce PanoVideoBench, the first manually annotated benchmark for comprehensive panoramic video understanding. Its 1,000 questions span five categories and 15 question types, with diagnostic annotations for target search and panoramic capabilities. Evaluation of 11 proprietary, open-source, and panoramic-specialized models identifies cross-view tracking and sphere-time evidence integration as their weakest capabilities. To strengthen these capabilities, we construct PanoVideoCoT, the first reasoning dataset grounded in real panoramic video, through sphere-time-grounded synthesis that converts verified panoramic observations and successful agentic explorations into 13,465 QA-CoT pairs. Starting from a general video MLLM, PanoVideoReasoner learns from this corpus through Panoramic CoT SFT and Evidence-Privileged On-Policy Self-Distillation, a panoramic instantiation of Vision-OPD's OPSD in which the teacher uses verified perspective views to guide student-generated trajectories. It reaches 55.7% on PanoVideoBench, improving the backbone by 12.5 points and the previous SOTA by 2.6. The gains concentrate on the same spatial, cross-view, and temporal capabilities exposed by the benchmark, showing that grounded sphere-time supervision can equip a general video MLLM with panoramic reasoning.
Abstract (translated)
URL
https://arxiv.org/abs/2610.42082
PDF
-
Robustness of Chatbot LLM Judgments of Presidential Actions and Hypothetical Reforms
Abstract
Chatbots are increasingly consulted about public affairs, and their judgments of policies are useful only if they are robust. We measure this robustness in the setting of an ordinary user, through consumer chat interfaces. For 30 days, five chatbot configurations (Mistral, Gemini Flash, Claude Sonnet, and Claude Opus at two effort settings) were asked twice a day to score the long-term effect on the United States of 40 actions of the last four presidential administrations and of 15 hypothetical reforms, each with a confidence and a short comment. The judgments are robust: all configurations agree on the sign and the ranking of the administrations' mean scores, and two different configurations asked on the same day give opposite signs to the same action in about 4% of cases. Responses from different days differ more than same-day responses (which shared the order of the items); the difference grows slowly over the month, except where the service itself changed. We identify three such changes, which shifted the judgments while leaving little or no trace in the interface. Treating the comments as a retrieval task, we find them specific to each item and consistent across runs and chatbots, with lexical and embedding similarities agreeing. We release the dataset of 16,500 judgments.
Abstract (translated)
URL
https://arxiv.org/abs/2610.39398
PDF
-
Same Question, Different Answers: Auditing Persona-Dependent Variation in Six LLMs on Smoking-Cessation Advice
2026-10-30 17:03:35Artin Yousefi, Aizierjiang Aiersilan, Lorien C Abroms, David A BroniatowskiAbstract
Hundreds of millions of people now ask large language model (LLM) chatbots for health advice, which raises a question that is simple to pose and hard to answer: does everyone get the same advice? We audit six widely deployed LLMs from four providers on smoking cessation, a domain with codified clini- cal guidance and substantial public-health stakes, by putting the same four cessation questions to twelve user personas that differ in life stage, self-described expertise, and smoking history, for a total of 2,880 responses. What the models say depends on who appears to be asking. Every model named FDA-approved prescription cessation medications less often to teen personas than to adult personas, by 32 to 67 percentage points, and in five of six models a persona presenting health- care expertise received fewer facts than a persona presenting none. We then evaluate one deployable intervention: a short system prompt specifying a factual floor, the set of core med- ical facts every user should receive when they are relevant, while leaving tone, vocabulary, and framing free to adapt. The prompt narrows persona differences in every model, but its effect on overall fact coverage varies in both magnitude and direction across models, so a prompt validated on one system cannot be assumed to transfer to another. Two cautions frame these results. Equal mention rates are not always the correct clinical target, because prescription cessation pharmacother- apy is not recommended for minors, so a defensible factual floor must be age-aware rather than uniform. And because our personas vary on several attributes at once, we report scenario- level differences rather than isolated effects of age or expertise. We release all 2,880 responses and the analysis code.
Abstract (translated)
URL
https://arxiv.org/abs/2610.42039
PDF
-
Latent-Lens: Visual Perception in Small Language Models Through Latent Communication
2026-10-30 13:53:44Mo Shahdloo, Faegheh Sardari, Mathew SalvarisAbstract
Bridging independent vision-language models (VLMs) and small language models (SLMs) in a modular multi-agent system typically forces them to exchange information as discrete text, which creates a lossy bottleneck for spatial and fine-grained details. This constraint is especially acute for SLMs, whose limited capacity makes them more reliant on the quality of the input signal. Recent latent-communication methods exchange states between language models. How a VLM can pass visual information to a separately pretrained text-only SLM though remains unexplored. We propose \textbf{Latent-Lens}, a lightweight codec that bridges the representation gap between a frozen VLM and a frozen SLM directly in latent space, enabling them to communicate purely through hidden states. The codec consists of a domain encoder that projects hidden states from the VLM into the SLM's embedding space, and an attention decoder formalised as a LoRA that teaches the SLM to attend to these visual prefixes, while training only ${\sim}1.6$M parameters (0.2\% of the combined model size). On visual question answering (CLEVR, GQA), Latent-Lens achieves $81.0\pm7.9$\% and $65.4\pm5.4$\%, compared with $58.3\pm5.7$\% and $61.8\pm1.0$\% for text communication with the same LoRA budget (+30.2--69.8pp relative to frozen text communication), while on image captioning (Flickr8k) it improves over the same baseline by +16.5 (+18.8 frozen). We also evaluate Latent-Lens with TinyLlama-1.1B and InternVL2-1B, which use different architectures, vocabularies, and hidden dimensions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.41751
PDF
-
Nereus: A Large-Scale Underwater Dataset for Fine-Grained Attribute Understanding and Grounded Counting Perception
2026-10-30 13:08:23Jingtao Deng, Zihao Jin, Yuehe Chen, Feixiang He, Shuiwang Li, Dan ZengAbstract
Underwater imagery records rich ecological information and supports applications such as science communication, fishery monitoring, ecological conservation, and oceanographic research. Beyond general scene understanding, these applications require models to recognize fine-grained biological traits and reason over spatially grounded evidence. However, existing underwater datasets often provide coarse category labels, generic captions, or total-count supervision, which limits their use for organism-level analysis and ecological reasoning. We introduce Nereus, a large-scale dataset for Underwater Fine-grained Perception that jointly supports Fine-Grained Object Attribute Understanding and Grounded Counting Perception. Nereus contains 88K images and 2.68M question-answer pairs, including annotations for 136K marine organism objects across 2,690 species and 1,515 genera, as well as 65K point annotations and 67K grounded counting QA pairs. Experiments with representative multimodal models show that Nereus improves downstream adaptation over existing underwater-domain data and provides a foundation for future research on fine-grained marine perception.
Abstract (translated)
URL
https://arxiv.org/abs/2610.41705
PDF
-
GraphNIAH: A Benchmark for Long-Context Reasoning over Real-World Graphs
2026-10-30 13:07:47Minh-Nam Tran, Abulhair SaparovAbstract
The needle-in-a-haystack (NIAH) benchmark was introduced as a test for measuring a model's ability to find small relevant facts within a very large body of input text. However, existing NIAH benchmarks each suffer from one or more of the following limitations: (1) reliance on synthetic data, so their evaluations may not generalize to real-world settings; (2) reuse of previously published datasets, leading to the risk of benchmark contamination; (3) dependence on manual annotation, which requires human effort to create and update data and is prone to future benchmark contamination; and (4) unrealistic simplification of long-context multi-hop reasoning where each next step is trivial given previous steps, while real-world reasoning often requires planning over multiple possible proof paths. To address these limitations, we introduce \Benchmark, a dynamic benchmark built from real citation relationships among scientific papers published after the evaluated models' knowledge cutoff date, which utilizes real-world information while ensuring no benchmark contamination. The fully automated benchmark construction pipeline samples representative subgraphs of the citation graph formed by these papers and converts each subgraph into a long-context QA example with a verifiable answer. We use GraphNIAH to evaluate eight LLMs from the GPT-5.6, Gemini-3, and Qwen3.5 families. We find that accuracy falls as questions require longer reasoning paths: at question depths of 16 and 32, the GPT-5.6 and Qwen3.5 models answer YES/NO questions no better than random guessing, mostly by answering NO; four of the eight models answer no counting example correctly, and the others model get worse as the subgraph size increases. These findings suggest that current LLMs often stop searching the citation relationships in their context before they reach the answer, even though every fact they need is in the context.
Abstract (translated)
URL
https://arxiv.org/abs/2610.41704
PDF
-
Keepace.ai: A Technical Report on a Large Language Model for Sports and Health
2026-10-30 12:29:47ai Team, Bowen Wu, Heng Wang, Jufeng Mi, Kang Liu, Kun Li, Ling Man, Long He, Suli Fan, Xiaosen Jiang, Yi Wang, Ying He, Yu Tang, Yunhan Li, Zifang LiuAbstract
Large language models hold promise for delivering safe, scientifically grounded, and personalized exercise guidance at scale, yet the sports-and-health domain still lacks curated training corpora and rigorous evaluation benchmarks. Keepace (this http URL) is a domain-specialized LLM built on Qwen3.5-27B through post-training, currently optimized for Chinese-language exercise guidance across three tasks: workout plan generation, sports-and-health knowledge QA, and personal sports data analysis. In internal rubric-based evaluations designed with domain experts, Keepace outperforms its base model and larger general-purpose baselines, with the most pronounced improvements in exercise safety, boundary compliance, structured output completeness, and scientific accuracy. This report summarizes the current design, data pipeline, safety-first evaluation framework, and evaluation results of Keepace.
Abstract (translated)
URL
https://arxiv.org/abs/2610.38249
PDF
-
When Depth Lies: Benchmarking Vision-Language Models on Mirror-Induced RGB-D Ambiguity
2026-10-30 12:25:32Hao Yin, Tianchen Guo, Heming Du, Yan Ke, Yanbin Liu, Xin YuAbstract
Vision-language models (VLMs) have made significant progress in spatial reasoning, particularly by integrating RGB images with depth data, which is often assumed to faithfully reflect physical distances. However, mirrors violate this assumption: they distort depth readings while reproducing object appearances, making reflections difficult to distinguish from real objects. As existing benchmarks rarely consider such reflective scenarios, VLM performance in mirror scenes remains largely unexplored. In this work, we introduce a mirror-centric RGB-D dataset from reflective scenes to evaluate the spatial reasoning capabilities of state-of-the-art VLMs in these scenarios. Specifically, we collect 1,011 RGB-D images from 525 real-world scenes, spanning diverse indoor and outdoor environments, six reflective surface types, and both daytime and nighttime conditions. All images are annotated with reflective surface region masks and mask annotations for both directly observed and mirror-reflected objects. Then, we curate a benchmark named MIRA-Bench (Mirror-Induced Reasoning Ambiguity Benchmark), comprising 14,140 QA pairs across eight sub-tasks organized into three cognitive levels: Reflection-aware Perception, Spatial Reasoning, and Scene-level Decision-making. We evaluate state-of-the-art VLMs on MIRA-Bench and find consistent failures across tasks: models misclassify whether objects are directly observed or reflected, fail to match reflected instances to their real-world counterparts, and produce erroneous spatial inferences. Remarkably, providing depth data actually degrades reasoning performance, implying that VLMs cannot identify incorrect depth in mirror scenes. The code and data are available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.41651
PDF
