Chat
-
False Claims, Credible Images: A Red-Teaming Benchmark for Commercial Image Generators
2026-10-08 02:35:39Zeyu Ye, Yanchun Li, Sibei He, Meng Xie, Hangtao Zhang, Xianlong Wang, Li Zeng, Jiahao Chen, Yichen Wang, Junhui Wang, Ziqi ZhouAbstract
Image-generation models can now produce text-rich, natural-looking visual artifacts that are hard to distinguish from real-world evidence, such as news reports and textbook pages. Yet, the same capability introduces a new risk: these models can just as easily fabricate visual misinformation. Even commercial models (e.g., GPT-Image-2) readily produce it. Curiously, we find that these models can recognize a claim as false when asked, yet still render that very claim as credible visual evidence. This discrepancy points to a blind spot in current alignment: safeguards judge what an image shows, not what it asserts; however, existing red-teaming benchmarks target conventional harmful content, such as violent or explicit imagery, and say little about where the alignment boundaries lie for visual misinformation, especially in commercial models. To fill this gap, we introduce EpiReal-Bench, the first systematic benchmark for evaluating visual misinformation risks in commercial image generators, comprising 10k false-claim prompts and 10k corresponding generated images that span 10 real-world claim categories and 10 credible visual formats. We further introduce EpiReal-Attack, a skill-guided black-box optimization framework that uses Pareto-based selection and multimodal feedback to identify commands that bypass alignment safeguards while preserving visual realism, textual legibility, and semantic fidelity. Experiments on four commercial models reveal that more than 70% of false-claim prompts elicit images that faithfully depict the corresponding misinformation, and EpiReal-Attack pushes this rate to 95%. Most worryingly, these models are only a click away, and their outputs are cheap to spread yet hard to disbelieve, leaving this dimension of alignment largely unguarded.
Abstract (translated)
URL
https://arxiv.org/abs/2610.11112
PDF
-
Using LMs to Model the Effects of Context and Coreference during Sentence Comprehension
2026-10-08 01:48:04Kohei Kajikawa, Lin Ai, Tatsuki Kuribayashi, Ethan Gotlieb WilcoxAbstract
Language models (LMs) are often used as a tool to model human language processing. Recent studies suggest that severely restricting LMs' context window improves their fit to human psycholinguistic data by simulating human working memory constraints. However, it is possible that this strict memory-decay approach overlooks humans' reliance on long-range structural representations, such as discourse structre. In this work, we systematically vary the context window size of GPT-2 across four large-scale naturalistic English reading-time datasets and observe a U-shaped relationship: Although restricted contexts (< 20 tokens) successfully capture local memory limitations, expanded contexts (500--1,000 tokens) ultimately yield the highest overall psycholinguistic fit. To investigate the mechanism driving this benefit, we conduct a counterfactual inference-time experiment that disrupts cross-sentential entity chains by pronominalizing repeated discourse entities. Obscuring these structural linkages significantly degrades the predictive power of larger context windows by 20% to 40%. Our experiments demonstrate that tracking long-range coreference relations is one important factor for the alignment between LM surprisal and human reading behavior, and approximate the extent to which human comprehenders use global discourse relations during language processing.
Abstract (translated)
URL
https://arxiv.org/abs/2609.32119
PDF
-
AgentHorizon: Evaluating Agentic Judges for Long-Horizon Computer-Use Tasks
2026-10-08 01:03:48Xing Han L\`u, Dheeraj Vattikonda, Sina Hajimiri, Fatemeh Pesaran Zadeh, Parishad BehnamGhader, Ghazwa Darwiche, Amirhossein Kazemnejad, Christopher Pal, Alexandre Drouin, Siva ReddyAbstract
Computer-use agents are capable of completing complex tasks, increasing the use of automatic judges to determine success, either for training or for evaluation without human involvement. Despite their flexibility, their reliability on long tasks spanning multiple applications remains unclear. A trajectory, composed of long sequences of screenshots and actions, may appear complete, but in reality violates constraints from the instruction or introduces an unwanted side effect. To identify these errors, a judge needs to examine the trajectory with respect to the user's instruction. To this end, we introduce AgentHorizon, a benchmark of 1,373 computer-use tasks (instruction-trajectory pairs) drawn from 166 hours of human-recorded trajectories spanning three operating systems. By recording trajectories for closely related instructions, we can construct negative tasks by swapping the instructions. This paired design evaluates judges on their ability to distinguish a successful trajectory from one that completed a similar (but incompatible) request. We release the benchmark under three splits: a frontier split, AgentHorizon (AH), a simplified split, AgentHorizon-Simple (AH-S), and a development split, AgentHorizon-Development (AH-D). We further evaluate eleven judges by (1) directly passing the full trajectory (with up to 300 screenshots and actions), and (2) using them as coding agents across five agent harnesses. We find that our best agentic judge, GPT-5.5, achieves 80.9% balanced accuracy on the AH subset. We find that tool-use improves certain models but results in worse performance for open-weight models, and that judges differ drastically in their ability to accept a valid trajectory and reject failed ones. Our findings highlight the need for judges that are capable of locating and verifying often hidden evidence that a task was properly completed inside long interaction histories.
Abstract (translated)
URL
https://arxiv.org/abs/2610.11050
PDF
-
Black-Box Forensics for Conversational LLM Agents
2026-10-07 23:46:14Isadora White, Yasaman Jafari, Taylor Berg-KirkpatrickAbstract
As LLM-powered scams proliferate, black-box forensics for conversational LLM agents offers a path to accountability for systems hidden behind anonymous endpoints. Identifying the base model behind a chatbot endpoint (attribution), without model parameter access or knowledge of the hidden system prompt, would let investigators trace AI-enabled scams back to the providers whose models power them. Detecting when two endpoints run the exact same system prompt (fingerprinting), even one novel and unseen, would link individual scams into criminal networks and expose silent API changes. We conduct an empirical investigation of both capabilities. Our attribution classifiers identify the base model behind an agent with 98% accuracy from a few turns of non-adversarial conversation. Attribution of system prompts, while possible, requires retraining on a large amount of data for each prompt; system prompts in the wild are unbounded and ever-changing, making this approach costly. To tackle this more open-ended setting, our cross-encoder fingerprinting method achieves an AUC of 0.768 and an F1 of 0.703 on entirely unseen system prompts, and aggregating 50 interaction conversations from each target agent boosts AUC to 0.943. Conversational agents with unseen system prompts can thus be fingerprinted with robust accuracy from a few turns of ordinary conversation.
Abstract (translated)
URL
https://arxiv.org/abs/2606.22698
PDF
-
Loose Lips: Understanding Privacy Leakage in Agentic Dialogue
2026-10-07 23:25:11Soumya Sharma, Ayana Hussain, Heber Hwang Arcolezi, Ulrich A\"ivodji, Spandana Gella, Adriana Romero-Soriano, Golnoosh FarnadiAbstract
As LLM agents are increasingly deployed in high-stakes enterprise settings, understanding how they handle sensitive information in agentic dialogue has become critical. Privacy leakage in such settings requires reasoning about whether information flows respect the contextual norms under which they originated. We extend CIMemories to CIMEMORIES DIALOGUE, a benchmark of 4,410 two-agent dialogue instances grounded in Contextual Integrity theory. Using this benchmark, we systematically evaluate 7 open-weight models from the Qwen3 and Llama 3.1 families across conversation lengths of 1 and 6 turns under three prompting conditions. We find that task capability and privacy leakage are fundamentally coupled: more capable models are systematically leakier, with no model in our evaluation able to decouple the two. Thinking models consistently leak more than their non-thinking counterparts, a pattern we attribute to the extended reasoning surface created by chain-of-thought. Leakage compounds with conversation length, rising up to 3x from 1 to 6 turns. Our results suggest that the coupling between task capability and privacy leakage is not an artifact of misconfiguration but a structural feature of current agentic systems.
Abstract (translated)
URL
https://arxiv.org/abs/2610.14071
PDF
-
Omni-Diffusion-Distill: Few-Step Distillation of Unified Multimodal Diffusion Large Language Models
2026-10-07 23:24:16Hong Huang, Chenhongyi Yang, Junzhe Sun, Animesh Sinha, Wuyang Chen, Yifan JiangAbstract
Unified multimodal diffusion large language models (dLLMs) offer a single architecture for both image generation and multimodal understanding, but their iterative decoding requires tens to hundreds of forward passes. Existing few-step distillation methods largely focus on either image generation or text generation, making it unclear how to compress a fully discrete multimodal dLLM into a single efficient student while preserving both generation and understanding. We introduce Omni-Diffusion-Distill, a unified two-stage distillation framework that retains strong generation and understanding capabilities while substantially reducing the inference cost of a unified multimodal dLLM. Omni-Diffusion-Distill aligns the distillation of both generation and understanding, for both images and text, in the discrete token space. In the first stage, the student is trained to skip decoding steps by replaying cached teacher trajectories, and in the second stage the student is refined on intermediate states along its own rollouts. We further remedy two sources of degradation in unified distillation with a pairwise collision penalty that reduces repetition under parallel text decoding, and entropy-matched guidance that prevents entropy collapse caused by fitting the sharpened teacher distribution in image generation. Omni-Diffusion-Distill achieves state-of-the-art trade-offs between decoding efficiency and generation and understanding performance for multimodal dLLMs, reducing image generation from 128 to 8 decoding steps and multimodal understanding from 512 to 64, giving 18.2x and 21.2x wall-clock speedups. Under these budgets, it scores 0.828 on GenEval and 83.0 on DPG-Bench for text-to-image generation, while reaching GPT judge scores of 20.0 on MM-Vet and 57.2 on COCO captioning (twice the teacher's 28.4 at the same steps) for multimodal understanding.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10990
PDF
-
Grammar Concept Annotation at Scale: Deployed Fine-Tuned Small Language Models Outperform Prompted Frontier Models
2026-10-07 19:31:24Marjan Celikik, Ana Peleteiro Ramallo, Javier MoralesAbstract
Corrective feedback is among the best-evidenced drivers of second-language acquisition, yet corrections delivered during lessons rarely accumulate into an actionable view of grammar mastery. Prompted frontier models can provide such a view from learner--tutor lesson transcripts, but they are costly at scale. We close this gap by fine-tuning Qwen3.5 small language models (SLMs) on filtered and rebalanced teacher-generated supervision, then deploying an efficient 0.8B model in an end-to-end grammar mastery tracker for all English learners on our platform. Internalizing the annotation contract into adapter weights enables pairing the 0.8B model with a compact matched prompt rather than verbose instructions. On two human-curated benchmarks, both the deployed 0.8B model and a 4B reference comparator outperform prompted GPT-5.4 and GPT-5.6 Sol in precision and recall under nested matching criteria of increasing strictness: concept, evidence span, and correctness. The deployed 0.8B SLM reduces serving cost by approximately 16$\times$. A feature-level online experiment shows significant gains in learner engagement ($+15.8\%$) and key business metrics, including scheduled hours ($+2.1\%$) and GMV from new lessons ($+13.2\%$).
Abstract (translated)
URL
https://arxiv.org/abs/2610.10827
PDF
-
NavGPT-3: Harnessing Context in a Hierarchical Navigation Runtime
2026-10-07 18:45:56Gengze Zhou, Yicong Hong, Jiazhao Zhang, Xunyi Zhao, Jian Zhou, Zixing Lei, Zun Wang, Chongyang Zhao, Xionghui Chen, Stephen Gould, Anton van den Hengel, Qi WuAbstract
Language models trained with long-horizon agentic reinforcement learning can generalize knowledge through reasoning, express precise actions, and pursue goals over many steps, raising the ceiling on what an embodied agent can understand and decide. Physical interaction, however, remains the domain of action policies, which provide dense, low-latency control. We present NavGPT-3, a harness that connects the two models, with an OS-like runtime built above it: reasoning, acting, and monitoring run as threads with their own context, tools, and permissions, while the runtime schedules them and decides which thread controls the robot's motion, so that the robot can react to sudden real-world events through interruption and thread switching. Beneath it, our action policy NavGPT VLA, trained on 19.28M examples, allocates visual tokens using codec allocation, in proportion to scene change; its 8B model alone reaches 74.51 SR on R2R-CE and leads RxR-CE with 78.19 SR. With the complete harness, NavGPT-3 sets the state of the art on R2R-CE (81.51 SR) and, for the first time, brings an autonomous agent to human level: on RxR-CE it matches human followers in success (90.43 vs. 90.4 SR) and path fidelity (78.47 vs. 77.7 nDTW) at 1 min 22 s per episode, versus roughly 3 min for a human. We comprehensively ablate the harness design and the interaction between the two models, showing how tools and the action policy shape the path from language-model reasoning to physical control: when NavGPT VLA executes the route, the reasoning loop shortens and the system's minimum reaction time falls from 3-19 s per language-model decision to 0.5-1 s per action-policy step (1-2 Hz). These results show that designing this embodied interface is central to connecting frontier language-model intelligence with low-level physical control. We will release all models, code, and evaluation records.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10787
PDF
-
Clarify, Then Focus: Statement Normalization for Conversation Analytics at Scale
2026-10-07 18:22:55Mikhail L. Arbuzov (Independent researcher), Karan Dave (Independent researcher), Evgeniya Dontsova (Independent researcher), Yaodong Hu (Independent researcher), Vincent Lao (Independent researcher), Navita Jain (Independent researcher), Sisong Bei (Independent researcher), Dmitry Dimov (Independent researcher)Abstract
Enterprise conversation analytics asks many questions of millions of interactions. Each question can require reconstructing what people mean and identifying which information matters, repeating costly interpretive work across the same transcripts. We propose a simple principle: clarify the text, then focus the reader. Statement normalization transforms dialogue into short, speaker-attributed statements with source references and semantic tags. The statements make meaning more explicit; the tags support selecting evidence for a particular question. Downstream models can use the full representation or a relevant subset, depending on what helps them make the decision. In an offer-suppression task on customer-service calls, normalization improves a supervised classifier without selection, while weaker prompted readers benefit from both normalization and selection. A small model can learn the normalization contract, while lightweight encoders handle tagging and downstream decisions. Sharing this preparation across questions supports an inference pipeline built entirely from small models, making analytics over millions of conversations substantially less expensive.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10758
PDF
-
Conversational Task Disambiguation over Tabular Data: Leakage-Aware Formulation, Benchmark Suite, and Training
2026-10-07 18:09:55Nafiseh Ghoroghchian, Luis Scoccola, Tina Sedaghat, Omid Vaheb, Hannah Chen, Dino D'Agostino, Keyvan GolestanAbstract
Conversational task disambiguation over tabular data uses dialogue to resolve missing information about a user's intended task before producing a solution over tables or databases. Existing evaluation and training lack a leakage-aware foundation. Task success mixes the agent's disambiguation and solution-generation capabilities and can also reflect oracle leakage, that is, information that a user simulator reveals beyond what a real user would. Existing datasets also lack a shared representation of ambiguities and access boundaries. We introduce the notion of an ambiguous verifiable task, which formalizes ambiguities and resolutions, decomposing the agent into an asking policy and a solution policy, and the environment into an oracle and verifier. This framework provides baselines and metrics for evaluating task disambiguation separately from solution generation, formal definitions of oracle leakage, judge-free leakage diagnostics, and a training objective for the asking policy. We instantiate the framework in text-to-SQL with AmbiTab, a benchmark suite that unifies six ambiguous datasets under a common representation specifying what the agent, oracle, and verifier may access. We evaluate clarification strategies and oracle leakage, and train an asking policy with reinforcement learning. The trained asker improves our disambiguation metrics on all six datasets and task success on five, and our leakage diagnostics measure how training affects oracle leakage.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10740
PDF
-
QuadTok: Quadtree Visual Tokenizer for Autoregressive Image Generation
2026-10-07 17:47:42Yucheng Mao, Zeyuan Chen, Xiaojun Shan, Xiang Zhang, Divyansh Srivastava, Bingnan Li, Zhuowen TuAbstract
We introduce QuadTok, a novel framework for visual tokenization and autoregressive image generation. Compared to traditional approaches using 2D grids or 1D token sequences, we propose a hierarchical quadtree structure, bridging the gap between 2D spatial binding and 1D sequence-level flexibility. The QuadTok tokenizer dynamically allocates representational capacity to visually intricate areas while leaving homogeneous regions at a coarse resolution. Compared with a fixed 256-token grid, our ImageNet-trained tokenizer saves approximately 10% of tokens on ImageNet and 9% when transferred zero-shot to the COCO dataset, while maintaining comparable reconstruction fidelity. Furthermore, the natural causality introduced by the tree structure seamlessly enables autoregressive image generation. Conditioned on a quadtree topology supplied before generation, our 947M GPT-style generative model achieves a 2.08 gFID on the ImageNet $256 \times 256$ benchmark. Additionally, leveraging the strong spatial correlation preserved by the quadtree structure, the QuadTok generator enables zero-shot spatially controlled image generation capabilities. Code: this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10497
PDF
-
Training Parallel Speculative Draft Models by Directly Minimizing Expected Decoding Rounds
2026-10-07 16:56:29Yunxiao Zhao, Changxiao CaiAbstract
Speculative decoding accelerates large language model inference by using a low-cost draft model to propose tokens that the full-size target model verifies in parallel. Parallel and semi-autoregressive (semi- AR) drafters improve drafting efficiency by proposing an entire block in a single forward pass, but training them raises a new difficulty: the draft distribution for a given position depends on where the decoding round starts, and where rounds start depends on how many tokens earlier rounds accepted. Existing training objectives typically rely on block-local surrogates that ignore this cross-round coupling, and therefore do not directly optimize the global decoding efficiency. In this work, we develop a theoretical framework for training and evaluating these drafters by representing speculative decoding as a Markov reward process. This formulation yields the Expected Decoding Rounds (EDR) objective, which weights local rejection costs by state occupancies and exactly equals the expected number of decoding rounds. Unlike prior surrogate objectives, EDR introduces no auxiliary hyperparameters. We then derive an exact temporal-difference gradient that supports unbiased stochastic optimization from target-model rollouts. The same framework also yields an exact offline evaluator for round counts, enabling paired drafter comparisons on shared target rollouts without running speculative decoding. Finetuning two state-of-the- art drafters, DSpark and DFly, with EDR consistently improves mean accepted length and outperforms existing training objectives across nine benchmarks spanning math reasoning, code generation, and chat.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10411
PDF
-
Nobody Truly Agrees on Sentiment: Humans, Bespoke Tools, and LLMs Struggle with Social Media Texts
2026-10-07 16:11:12Himarsha R. Jayanetti, Sivakanesan Dhanushkanda, Shuai Hao, Michael L. Nelson, Michele C. WeigleAbstract
Social media is a rich source of real-time public sentiment, but widely used sentiment analysis tools are often applied without understanding their limitations. In this study, we evaluate the inter-rater reliability of three bespoke sentiment analysis tools (TextBlob, VADER, and Twitter-roBERTa-base) and three large language models (LLMs: Qwen3-32B, GPT-OSS-120B, Llama-4-Maverick-17B) against six human raters across 100 tweets. We measured agreement using two statistical measures: Cohen's kappa for pairwise comparisons and Fleiss' kappa for multiple raters. Even among the human raters, our results showed only fair agreement, highlighting the subjectivity of sentiment analysis. Higher agreement was observed under the binary sentiment classification (negative vs. non-negative and positive vs. non-positive) than under the three-class classification across both humans and automated tools. The Twitter-roBERTa-base model showed the strongest alignment with human ratings, outperforming both bespoke sentiment tools and LLMs, particularly in distinguishing negative versus non-negative sentiment. LLMs showed substantial agreement among themselves and moderate to substantial alignment with humans, performing better in positive vs. non-positive classifications. Our findings underscore that domain-specific fine-tuning remains crucial for reliable social media sentiment analysis, and human-centered evaluation remains essential for establishing gold-standard labels.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10318
PDF
-
CARES: A Controlled Synthetic Benchmark of Speaker Reactions to Sound
2026-10-07 15:05:59Marcel Gibier, Thomas Thebaud, Olivier Bo\"effard, Jean-Fran\c{c}ois BonastreAbstract
Automatic audio scene description turns a recording into a text account of a situation. One difficulty is deciding which elements of the audio should be kept, since a description cannot include them all. Annotators disagree about this, making a ground truth hard to obtain. In this work, we first define the ground truth, then generate the data. We focus on audio events and define sound salience with a simple rule: a sound is salient when a speaker audibly reacts to it. For scale and variety, a controlled set of scenarios fixes the ground truth, and a language model writes the dialogues. The resulting corpus, CARES, contains 10,000 two-speaker scenes. We then benchmark six audio-language models on three tasks: identifying the scene, tagging the sounds present, and classifying reactions. We show that these models hear the sounds but miss how the speakers react to them.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10208
PDF
-
Beyond Anonymous Captions: Grounding Character Identity in Video Captioning and Question Answering
2026-10-07 14:41:29Anas Filali Razzouki, Killian Steunou, Khalil Guetari, Thomas Kling, Moun\^im El-Yacoubi, Yannis TevissenAbstract
Linking people's appearance and actions to character identities is essential for understanding video narratives. We present a framework for identity-aware video captioning and person-centric question answering that combines automatic character identification, explicit spatial grounding, and task-specific adaptation. Starting from LSMDC v2 movie clips, our pipeline matches detected faces to actor reference images, tracks characters across frames, and builds inputs with identity-linked bounding boxes. A strong vision-language model generates identity-aware captions and questions, which are manually verified and filtered to create a benchmark of 750 captioned clips and 3,000 person-centric questions. We study five grounding strategies combining textual coordinates with visual face or estimated person boxes across Video-MLLM families at roughly 2B, 4B, and 8B parameters and larger frontier models. Combining visual face boxes with textual coordinates yields the most consistent performance across scales and significantly improves overall performance over coordinates alone. Smaller models tend to over-assign known identities when the queried person is not grounded, while larger models better recognize such UNIDENTIFIED cases. We introduce BAC by LoRA fine-tuning Qwen models at 2B, 4B, and 8B scales on about 32K identity-aware captioned clips. Across all scales, BAC outperforms every other evaluated model family of comparable size. BAC-8B reaches 93.20\% overall QA accuracy, ranking behind only GPT-5.6 Sol among the frontier models evaluated in our study. Overall, explicitly communicating who is where, together with lightweight task-specific adaptation, substantially improves identity-aware video understanding without changing the underlying architecture. We release the benchmark, training data, code, and BAC checkpoints at this https URL.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10163
PDF
-
I would rather quit NLP than read another paper like this: The rise of antithesis in NLP papers
2026-10-07 13:50:50Olga Zamaraeva, Adri\'an Gude, Roi Santos-R\'ios, Carlos G\'omez-Rodr\'iguezAbstract
For better or worse, LLMs are by now used routinely for scientific writing.\footnote{This paper is no exception; we did use AI to assist with writing some of the sections (see Acknowledgments).} Many have noticed that recent models fill papers with unnecessary antithesis, stating over and over what the work does not do, in ways that do not contribute to its precision or quality of expression and annoy reviewers \emph{rather than impressing them}. We study the construction \emph{rather than} in ACL papers from 2019, ACL-style arXiv papers from 2026, and papers written by GPT models from the same titles and abstracts. Its rate in 2026 is seven times the 2019 rate, and higher still in the GPT papers. Two annotators, blind to the source, find almost no 2019 use \emph{annoying} and about one in ten 2026 uses; they seldom agree on which, yet about half of 2026 papers contain a use that annoys each of them. \emph{Annoying} uses present the rejected alternative less favorably than legitimate uses. Raters of preference data and open reward models favor the construction, and an instruction to be honest promotes it. We conjecture that it is a side effect of post-training on pairwise preferences, which credit a disavowal in a single response and cannot register its cost across a text.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10092
PDF
-
From Pixel to Coding: Evaluating the Figure Reproduction Capabilities of MLLMs
2026-10-07 13:33:04Zijian Chen, Zhengyu Chen, Bohan Liang, Lirong Deng, Yushuo Zheng, Yanwei Jiang, Qi Jia, Kaiwei Zhang, Wenjun Zhang, Guangtao ZhaiAbstract
Multimodal Large Language Models (MLLMs) have demonstrated impressive capabilities in both visual understanding and code generation. However, existing benchmarks typically evaluate these two modalities in isolation, lacking a dedicated assessment of their unification, i.e., how a model can perceive complex visual structures and synthesize them into precise, executable code. Moreover, current visual code generation benchmarks often rely on simplified layouts within single programming environments, falling short of evaluating true unified multimodal reasoning. To bridge this gap, we propose FigCodeBench, a comprehensive framework for rigorously evaluating MLLMs on figure reproduction, integrating multimodal comprehension and generation. We first design a systematic dataset construction pipeline, resulting in a total of 6,194 instances that cover 7 functional categories and 4 types of programming languages. We further categorize figure reproduction into three tiers with visual and code complexity modeling, specifically targeting complex structural reasoning, varying aspect ratios, and dense geometric constraints. We introduce a multi-dimensional evaluation protocol, encompassing visual fidelity and syntactic isomorphism, that aligns highly with the Mean Machine Opinion Score (MMOS) and human preferences. Based on our framework, we conducted extensive experiments on 24 widely used proprietary and open-source MLLMs (e.g., Gemini 3.1 Pro, GPT-5.4, and Kimi-K2.5), where we observed a universal, non-linear performance cliff across different programming languages and difficulty scenarios for all models, and gained several insights, such as the significant metric decline in rigid declarative languages.
Abstract (translated)
URL
https://arxiv.org/abs/2610.10066
PDF
-
Sensitive-Topic Leakage Through LLM Routing Metadata: Measurement and Mitigation
Abstract
LLM routers pick a cheap or expensive model per request by its content, and many gateways and some cloud platforms can log that choice with content logging off. We measure this privacy channel beyond token counts, accounting for noisy labels and repeated prompts. We run pre-registered studies on 1.7 million real requests (WildChat-1M, LMSYS-Chat-1M) with two cost/quality routers and a domain router, survey eleven systems' logging, and test post-processing defenses. At matched length, the shift's direction depends on category and router. For RouteLLM at the 50% operating point, harassment and self-harm requests reach the strong model 19 points less often than comparable ones on prompts unseen in exploration, medical requests (exploratory: LLM labels failed their gate) 31 points less often on distinct prompts (both post hoc), and sexual requests 10 points more often (secondary); the other router's four are negative. Twenty RouteLLM decisions separate frequent medical askers with AUC 0.71, exploratory and below the pre-registered primary endpoint's 0.75 (domain router: 0.92, an upper estimate). Per-category length-matched parity with accurate labels removes the gap on real traffic, costing at most 0.2 accuracy points on RouterBench (post hoc), where routers' gaps on sensitive subjects (13-42 points, pre-registered) exceed those of an oracle routing by realized accuracy gain (1-11, post hoc). Per-conversation stickiness, per-user budget bands, and pooled parity fail, the last as categories' shifts differ in size or sign. A post hoc exact per-user rate hides only even-prefix strong counts and forfeits most self-assessed routing value; it preserves odd-position decisions, from which a post hoc log attack reaches AUC 0.73 after 20 RouteLLM requests (exploratory).
Abstract (translated)
URL
https://arxiv.org/abs/2610.09981
PDF
-
WorldBench: Evaluating LLMs on Three.js Voxel World Generation
Abstract
Large language models can now write complete, interactive 3D worlds as code, but grading those worlds automatically is unreliable. Existing judges take one view of the output: a vision-language model scores a few rendered snapshots, or a language model reads the source. On worlds written by five frontier models we find that the two views disagree on 32% of required items, mostly code that no frame shows, and that fixed views miss small close-up contents. We present WorldBench, a benchmark and judge for open-ended, LLM-generated this http URL worlds. From one prompt describing a floating voxel island with ten biomes, physics, and day/night and seasonal cycles, the judge explores the running world, controlling its clock, orbiting it, and sending a navigator agent to frame each biome, and reads the code for what it sees. Neither channel is trusted on its own: a code quote counts only if it is text the source contains, and visual claims are checked against measured pixels where the property is measurable. A mutation test, in which we remove features by construction, shows that code-only judging gives full credit to four of five removed features, because their code remains in the file. Our judge cuts the points kept on removed features by a third (5.44 to 3.55 of 7.11), and what it still credits is mostly code that exists but never runs. We evaluate five frontier models: Claude Fable 5.1, GPT-6 Astra, Kimi K3, Grok 4.7 and Gemini 3.1 Pro. Code, prompt, tests and judge configuration are available at this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.10622
PDF
-
RELATE: An Evaluation Framework for measuring Relational Orientation of Large Language Models
2026-10-07 07:10:10Shivam Shukla, Jihye Kim, Shubham Gaur, Mahnaz Roshanaei, Magy Seif El-NasrAbstract
Large language models (LLMs) are increasingly used for emotional support, raising concern that sustained use may draw users away from their real-world relationships. Yet existing evaluations primarily focus on the safety, empathy, or helpfulness of responses, leaving under-examined a relational question: where does the model orient the user for continued support? To address this question, we introduce relational orientation, a property operationalized through two non-exclusive dimensions: inward-facing (IF) language, which positions the AI as the user's ongoing source of support, and outward-scaffolding (OS) language, which encourages real-world human connection. Grounded in psychological and sociological literature, we formalize a taxonomy of relational orientation and present RELATE, a persona-conditioned framework for measuring inward-facing and outward-scaffolding language at the sentence level in multi-turn dialogues. RELATE pairs 76 help-seeking situations adapted from naturally occurring questions with three simulated user styles, providing 228 evaluation stimuli. In our experiments, we evaluate seven LLMs using dialogues with six assistant turns each, yielding 1,596 dialogues and 69,194 assistant sentences. We assess these sentences using a primary rubric-based LLM judge and apply a secondary judge to a subset. Under automated evaluation, we find that the proportion of sentences labeled as IF is higher at the sixth assistant turn than at the first, while the proportion labeled as OS is substantially lower for hesitant, indirect simulated users than for explicit, reassurance-seeking users. RELATE provides a reproducible framework and a sentence-level signal for auditing and steering the relational orientation of supportive LLMs.
Abstract (translated)
URL
https://arxiv.org/abs/2610.09569
PDF
