LLM
-
T-DIG: Text-Conditioned Diversity for Image Generation
2026-10-21 00:11:32Simran Khanuja, Rohit Gandikota, Gaurav Parmar, Graham NeubigAbstract
Text-to-image (T2I) platforms typically return several images per prompt so users can explore different visual interpretations and choose one that matches their intent. However, T2I models are usually optimized for single-image quality which comes at a cost of multi-image diversity, limiting this exploration. Diversity is also contextual: one user wants variations in background, while another wants them in artistic style, and the same user may later want variations in clothing. The generated images should differ along whatever axis the user cares about. In this paper, we introduce T-DIG, a text-conditioned diversity framework for image generation and editing. T-DIG is an iterative framework comprising multiple modules that lets users type prompts, inspect default generations, request variations using natural language, and continue refining outputs until they're satisfied. At its core, T-DIG uses an LLM policy that produces K axis-conditioned prompts or edit instructions for a frozen generator or editor. We train this policy with GRPO using a text-conditioned diversity reward and alignment penalties. Because the image generator remains frozen, T-DIG applies to open-weight, distilled, and proprietary models. We also release the first dataset for text-conditioned prompt diversity in image generation, with the training and validation splits built from human-written axes and a test set drawn from real user interactions with T-DIG. T-DIG far exceeds baselines in improving text-conditioned diversity, while preserving base-prompt alignment across a variety of metrics and human judgements. We show that user-directed diversity exploration can be added to a wide range of image generation models, open and closed, without the need to access their weights. Website: this https URL
Abstract (translated)
URL
https://arxiv.org/abs/2610.29003
PDF
-
FlashNext-RL: A System Optimization Study for Long-Horizon Agentic RL of Qwen3.8-Flash-Next
2026-10-20 23:53:09Zhewei Yao, Yuxiong HeAbstract
Long-horizon agentic reinforcement learning (RL) requires both efficient rollout generation and training on long sequences. However, the memory requirements and model-specific operators of Qwen3.8-Flash-Next make this difficult with existing open-source implementations. To resolve these limitations, we introduce FlashNext-RL, a model-specific RL training optimization framework based on prime-rl, with vLLM for software engineering (SWE) rollout generation. Our main focus is BF16 training of the full 48-layer model on a single 8xB200 node, with the first two layers, the n-gram table, the sparse-attention indexer and the multi-token prediction head frozen. We optimize host offload, gradient accumulation and the kernels for mixture-of-experts, sparse attention and the gated residual stream. In the full-model development study, non-padding training throughput increases from approximately 395 tokens/s/GPU in the initial 32K-token configuration to 4,438 tokens/s/GPU in the optimized 131K-token configuration, with hardware matmul throughput (aka without considering the cost of attention components) increasing from approximately 20 to 213 TFLOPs/GPU. A controlled cumulative study on an 8-layer model further quantifies the contributions of individual optimizations. Together with three inference nodes, the system supports a live SWE RL run with a target batch of 1,024 trainable rollouts and a 131,072-token context limit. Over steps 5-15, an iteration takes 22.4 minutes on average, with bounded entropy and trainer-sampler mismatch KL. These results highlight the importance of tailoring RL training to the model's architecture, memory layout and kernel requirements, rather than relying only on general-purpose RL infrastructure.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28998
PDF
-
Language Impairment as a Lens onto Large Language Models
2026-10-20 23:42:36Muxuan He, Wang Bill Zhu, Yue Chen, Elsi Kaiser, Weizhi Du, Yao Du, Samuel Nastase, Yanjun GaoAbstract
Although large language models (LLMs) encode language in highly distributed spaces, they remain vulnerable to focal disruptions---a phenomenon strikingly similar to how localized brain damage causes clinical aphasia in humans. To investigate this theoretical tension, we ask whether LLMs possess a functional structure analogous to the human language system, by introducing a novel mechanistic interpretability framework inspired by aphasia. Through two complementary approaches, classifier-guided lesioning and continuous activation steering, we systematically induce aphasia-like breakdowns in LLMs. We find that language functions can be successfully isolated to discrete structural circuits and continuous representational vectors, yet these functions remain densely entangled with significant circuit overlap and vector similarities across distinct aphasic profiles. Furthermore, while steering generalizes robustly across downstream tasks, structural lesioning remains asymmetrically constrained. By bridging neuropsychology and mechanistic interpretability, we demonstrate that LLMs, functionally analogous to biological brains, rely on a densely entangled architecture that can be systematically mapped through clinically-informed interventions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28992
PDF
-
PerMIT: Securing Persistent Memory for LLM Agents via Machine-Checked Integrity Theorems and Scalable Community Trust
2026-10-20 23:31:06Tanqiu Jiang, Min Bai, Sadat Shahriar, Wanyu Du, Sandesh SwamyAbstract
Persistent memory is a largely undefended surface for injection and poisoning, and the same risk limits memory sharing among agents. We present PerMIT, a memory architecture with provable integrity guarantees under an explicit trust assumption: only user-typed inputs and values confirmed at a gate are trusted. PerMIT mediates the release and use of untrusted environment-derived values through a finite memory interface, while preserving agent-driven multi-turn planning and tool choice. Over this finite interface, we prove two integrity theorems that are also machine-checked in Lean 4, which together ensure that environment-derived memory reaches a sensitive action only if confirmed at the gate and influences the agent's behavior only through what the gate displays. Every attack in our threat model thereby reduces to a single auditable decision, which a human, an LLM verifier, or both can make in practice. On Trojan Hippo, PerMIT records zero observed attack success with self-verification and lightweight verifiers. Because each confirmation is a logged verdict on exact content, endorsements accumulate across users, and sufficiently endorsed items become community-confirmed with no further verification required. Under explicit assumptions, we prove and corroborate in simulation a co-scaling property: as the community grows, the risks of promoting poison and of wrongly evicting benign knowledge fall, while per-user verification effort does not rise. On $\tau^2$-bench, communities of up to 90 autonomous agents use PerMIT to build shared knowledge bases with no direct memory injection promoted, and the resulting knowledge significantly improves new agents on held-out tasks, suggesting a path toward agents that safely self-improve at test time.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28987
PDF
-
Behavior Beyond Text: Evaluating Refusal Transfer Failure in LLM Agents
2026-10-20 23:14:30Ekagra Gupta, Raj Saha, Aniekan-Abasi Inyang, Adam Al GharibAbstract
While large language models are increasingly deployed as tool-calling agents, their safety evaluations still focus primarily on their ability to refuse harmful requests in text. We investigate whether the refusal behavior in text transfers to tool use. We construct a benchmark of 2,304 prompts across four domains, consisting of matched triplets that present the same harmful or benign request in text-only, standard tool-enabled, and adversarially framed tool-enabled settings. We then assess tool-call safety using 20 predefined forbidden actions. Testing five models of varying sizes and families, we find that requests refused in text frequently result in unsafe tool calls. Under standard tool-enabled prompting, this divergence occurs for 10-52 percent of harmful requests refused in chat. To mechanistically investigate this gap, we identify a refusal-associated direction in the residual stream of each of the five models. We find that in tool-enabled contexts, the activations along this direction decrease in four of five models. Furthermore, lower activation predicts unsafe tool calls across all five models (AUC 0.76-0.86). However, restoring the signal to the respective text-only value through activation patching only prevents a subset of unsafe calls, primarily by suppressing tool use rather than by producing safer calls. Steering along the refusal direction nearly eliminates unsafe calls only at strengths that also suppress benign tool use or degrade the output quality. Our findings indicate that the identified refusal direction contributes to whether a model acts, but only partially explains the refusal divergence. Therefore, evaluating the safety of tool-calling agents requires directly examining their actions along with their textual responses.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28982
PDF
-
Black-Box Red-Teaming of Multi-Agent Systems via Reinforcement Learning
2026-10-20 22:32:22Yuqi Jia, Hangsheng Zhang, Chenxi Liu, Junjie Liang, Xun ChenAbstract
Large language model (LLM) agents are increasingly deployed in multi-agent systems (MAS) to accomplish complex tasks. Prior black-box red-teaming attacks mainly focus on single agents, but we find that these methods are far less effective in MAS, where multiple sub-agents may not directly interact with the user. Therefore, we introduce ReMAS, the first reinforcement learning-based red-teaming framework tailored to MASs, which fine-tunes attacker LLMs to generate effective adversarial prompts for system prompt extraction. The framework follows a two-step process: first, a rewrite stage refines base attack prompts to increase extraction success, and second, a template generation stage constructs attack templates that improve the likelihood of invoking specific sub-agents and thus revealing their system prompts. Extensive experiments show that our method substantially improves attack success rates compared to existing black-box attacks, with transferability across different backend LLMs and MAS structures. These results underscore the vulnerabilities of multi-agent systems and the importance of developing stronger defenses against adaptive black-box adversaries.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28966
PDF
-
SLoSa: Small, Low-Power, and Safe Foundation Models for Robotic Task Planning
2026-10-20 22:09:44Weizhe Xu, Jialiang Fan, Essien Taylor, Zijun Wang, Russell Joseph, Fanxin KongAbstract
Safe robotic task planning requires planners that can reliably achieve task goals while satisfying safety constraints. Classical planners search over explicit symbolic task representations and can provide formal guarantees with respect to encoded safety constraints, but require rigid input formats. Their computational cost also increases rapidly with the search space. In contrast, large language models (LLMs) offer flexible task interfaces and strong generalization capabilities, but they are large, power-hungry, and pre-trained with little robot-specific safety data and knowledge. Thus, they can be difficult to deploy locally on robotic systems with limited resources. To address these limitations, we present SLoSa, a neuro-symbolic compression and fine-tuning framework for training foundation-model-based robotic task planners with Small size, Low power demand, and Safety awareness. The framework first constructs a safety-aware dataset using classical planners, an advanced teacher LLM, and formal verifiers, followed by two training stages: supervised fine-tuning and reinforcement learning (RL) fine-tuning guided by a verification-based reward machine to enforce safety alignment. We evaluate SLoSa across four Planning Domain Definition Language (PDDL) domains, two household simulation environments, local edge platforms, and a desktop robotic arm. Across the PDDL tasks, verifier-guided post-training raises the success rate of compact 4B-14B planners to 70.5-76.0%, while INT4 deployment on a Jetson Orin requires 17.85-26.41 W of average power. Results in household simulators and on the physical robot further demonstrate the applicability of SLoSa across diverse robotic task settings.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28955
PDF
-
Fooling LLM Safety Guards into Judging the Wrong Text
Abstract
Safety guards check content before an LLM application acts on it, and most paste that content into a plain-text prompt. Content that imitates the prompt can append a second task, and the guard answers that one instead. We call this target rebinding. Our explanation is that the guard grades the last complete task in its input, here a harmless decoy, so the harmful request goes through unchanged. In a pre-registered test, an appended forged query and decoy clear all 75 held-out AdvBench requests that Shieldstral-1.0-3B flags. The guard is not persuaded: a directive to ignore the text above clears none of them, and in a follow-up run a harmful decoy stops the attack. On our GuardRebind benchmark, each of five guards has an attack family, chosen per guard, that clears at least 93% of its flagged items. Rebinding explains this best on Shieldstral and gpt-oss-safeguard; on ShieldGemma and Llama Guard 3, generic evasion and trailing text respectively do part of the work, and Qwen3Guard falls mainly to a forgery of its whole task. Training the guard to score content no lower than its most harmful part cuts attack success from 75% to 0.8-1.6% on attack families, decoys and forged queries held out of training. All of these attacks append the forged structure after the request, and we added the held-out-decoy evaluation after a first round of test runs. Clean AUROC falls by at most 0.01, but our test of non-inferiority is inconclusive. A score-guided search still beats this trained guard on 47 of 50 items, usually by placing forged structure before the request. On about half of the items where the search wins, the same search also finds benign text, with no forged structure, that clears the guard.
Abstract (translated)
URL
https://arxiv.org/abs/2610.31695
PDF
-
Model Capacity Frontier: A Development Framework for LLM-Supported Social Simulation
2026-10-20 21:58:48Patrick Zhao, Yuxuan Li, Hirokazu ShiradoAbstract
Large language model (LLM) agents are increasingly incorporated into agent-based models (ABMs) on the premise that richer, context-sensitive human behavior can improve empirical fidelity. However, LLM-supported ABMs introduce an open-ended design space in which subtle modeling and configuration choices, such as prompt design and environment coupling, can substantially alter outcomes. Developing LLM-supported ABMs to simulate human collective behavior therefore requires navigating configuration sensitivity while identifying the limitations of the current model architecture. In this paper, we introduce the Model Capacity Frontier (MCF), a framework that characterizes the empirical performance attainable by a specified architecture across its admissible configurations. In multi-objective settings, this capacity is represented by an estimated Pareto frontier obtained through systematic configuration exploration. We demonstrate MCF in crowd simulation of university commencement events with approximately 12,000 agents, using empirical movement data on temporal occupancy, spatial distribution, and flow direction. We first characterize the capacity of a conventional crowd model and then evaluate successive LLM-supported architectural extensions. In the process, MCF reveals which dimensions of attainable performance expand and which remain limited, providing empirical targets for refinement and re-evaluation. Initial LLM integration improves flow direction but degrades temporal occupancy and spatial distribution; subsequent architectural refinement of the LLM layer improves temporal occupancy and recovers spatial performance, while retaining much of the earlier gain in flow direction. Our case study shows how MCF guides the systematic architectural development of LLM-supported ABMs in configuration-sensitive, open-ended design spaces.
Abstract (translated)
URL
https://arxiv.org/abs/2610.24168
PDF
-
Self-Improving Agents with World Model Predictive Paths
2026-10-20 21:46:29Haori Lu, Tianyu Gao, Yaoyao Liu, Yingying LiAbstract
Large language model (LLM) agents are increasingly expected to interact with environments and improve their decisions during deployment. However, how anticipating future outcomes and drawing on accumulated interaction experience can jointly support decision-making remains underexplored. We propose a test-time framework, inspired by Model Predictive Path Integral (MPPI) control, that combines multi-branch predictive analysis with experience-based adaptation. At each step, the agent proposes multiple candidate actions and imagines their outcomes with a world model, so that alternatives can be compared without being executed in the environment. Since language environments provide neither explicit dynamics nor costs, the predicted outcomes are evaluated for task progress and risk, and MPPI-style weights over the resulting costs guide the final decision. A dual-branch experience memory enables self-improvement: the planner branch refines candidate generation and action selection, while the world model branch improves next-observation prediction. Experiments on ALFWorld and ScienceWorld across three backbones demonstrate the strong overall performance of our method and the effectiveness of experience-guided predictive control. Our code will be made publicly available.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28945
PDF
-
Navigation Explainability for Object Avoidance with LLMs for Assistive Robots
2026-10-20 20:54:43Courtney Young, Yi Lian, Taery Kim, Sehoon HaAbstract
Blind and visually impaired individuals face uncertainty when navigating unfamiliar areas. Whether a cane or a guide dog is used for navigation, this may prove to be challenging. To mitigate this challenge, we present the Context-Aware Navigation Explanation framework that explains guide dog movements for obstacle avoidance and navigation using a large language model. This system integrates perceptual context with planned future trajectories to generate natural-language explanations of robot movement. We deploy the system on a Unitree Go2 platform in indoor navigation scenarios and evaluate how different explanation context configurations affect user trust and explanation satisfaction. Results show a strong positive relationship between explanation satisfaction and trust, and suggest that combining obstacle information with future trajectory leads to the most effective explanations. These findings highlight the potential of context-aware language explanations to improve transparency and trust in assistive robot navigation.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28920
PDF
-
DiffResearch: Accelerating Deep Research with Diffusion Language Models
2026-10-20 20:40:20Alan-Barsag Gazzaev, Alexey Gavrilov, Mikhail Mozikov, Sergey MuravyovAbstract
Deep research systems iteratively plan, gather evidence, synthesize, and revise, making end-to-end latency a key bottleneck for interactive use. Almost all production systems are built on autoregressive (AR) language models and inherit left-to-right decoding in planning and long-form writing. We present DiffResearch, an open multi-agent deep research framework that places a diffusion language model (dLLM) at the core of every stage, and measure how a diffusion backbone changes latency and report quality. Holding the agentic scaffold, retrieval code, prompts, and evaluation fixed, we swap the backbone between an open diffusion model and a comparable AR model. Across 100 Deep Research Bench tasks, the diffusion backbone is roughly 6 times faster end-to-end, driven by order-of-magnitude reductions in planning and writing. We further introduce dFast, a decomposition mode that reads intermediate denoising states and extracts subqueries before generation finishes, providing a 1.99 times median speedup for decomposition in isolation with no loss of report quality. On the benchmark's RACE rubric, dFast obtains the best overall score among the three configurations. Finally, in a public system-level leaderboard snapshot, an instance built on a stronger diffusion model ranks above several commercial systems. We release DiffResearch under an Apache-2.0 license.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28909
PDF
-
TrustAgent-RFF: Agentic Diagnostic Planning for Reliable RF Fingerprinting under Domain Shift
2026-10-20 20:14:53Chuan Liu, Tianya Zhao, Yiting Wang, Junqing Zhang, Xuyu WangAbstract
Radio frequency fingerprinting (RFF) models can achieve accuracy under source domain conditions but become unreliable when the receiver, distance, collection day, or environment changes. Incorrect predictions in unseen domains may receive high softmax confidence, making confidence thresholding insufficient for trustworthy deployment. We propose TrustAgent-RFF, a prediction level reliability framework for RFF that operates with a frozen in-phase and quadrature (I/Q) classifier without target domain labels. For each device prediction, the framework acquires radio frequency evidence: counterfactual stability, prototype consistency, acquisition domain similarity, and signal quality. A source calibrated trust estimator combines these diagnostics with classifier confidence, prediction margin, and evidence availability indicators. Source derived lower tail penalties and cross module interactions reduce trust when diagnostic signals indicate weak support. A large language model (LLM) based planner adaptively selects the next diagnostic action based on evidence, trust state, action costs, and remaining budget. A deterministic source calibrated decision layer updates the prediction state as Accept, Reject, or Defer. Experiments across five RFF datasets and classifier backbones reveal an average source to target accuracy drop of 38.7% and show that many incorrect target domain predictions remain highly confident.
Abstract (translated)
URL
https://arxiv.org/abs/2610.24930
PDF
-
SANE: Automated Scorer-Specific Metamorphic Testing for LLM Judges
Abstract
LLM judges increasingly mediate benchmark evaluation, model selection, data curation, and training. Their decisions are useful only when each judge responds reliably to changes affecting its scoring criterion. Existing methods for reliability testing perturb inputs with known expected effects on judge scores. Prominent LLM-judge testing harnesses combine human-authored perturbation recipes with synthetic test generation, leaving scorer-specific recipe ideation outside the automated pipeline. This limits their use to fixed perturbations and pre-assumed scorer types. We introduce SANE, an agent-based framework that encodes behavioral requirements for stochastic evaluators as reusable agent instructions. Given a scorer specification, implementation, and representative inputs, the pipeline first identifies scorer-specific capabilities, then automatically derives perturbation recipes and expected score relations for each. SANE also defines a scorer-independent validation framework for generated judge tests. It determines whether each recipe agrees with scorer intent and whether each instantiated case realizes the recipe and supports its expected relation. Because SANE generates a recipe inventory tailored to each scorer, the number of aligned recipes grows as scorers are added. Across five domains, SANE produces 3.7 times as many validated test families as the Judge Reliability Harness (JRH) at comparable end-to-end validity and detects scorer deficiencies in every domain, including professionalism and toxicity, where JRH detects none. On faithfulness, SANE achieves strong human alignment in judge ranking: its test-suite scores rank 11 judge models consistently with human-labeled FaithBench performance (Spearman rho = 0.852), while JRH does not (rho = 0.100).
Abstract (translated)
URL
https://arxiv.org/abs/2610.28886
PDF
-
DiaLong: Benchmarking Reminiscence in Long-Context Dialogues
2026-10-20 19:55:41Owen Bianchi, Samer Aslan, Ryunosuke SaitoAbstract
How well can transformer-based Large Language Models (LLMs) retain and recall memories in long contexts of dialogue? LLMs are known to forget long-range factual information in their context, yet the field lacks a comprehensive benchmark to quantitatively assess these shortcomings. We present DiaLong, a long-context dialogue dataset and accompanying memory task that standardizes evaluation in this space. It consists of 100 unique conversations created by humans and then augmented with LLMs, extending beyond lengths found in existing datasets. Our selected memory task is fact recall: classify 40 candidate facts about a dialogue as true or false. This task directly probes the model's ability to actively retain and retrieve information, serving as a critical measure of memory within extended conversational contexts. In testing both the latest LLMs and proposed memory enhancements on DiaLong, we demonstrate that more work needs to be done to enable LLMs with better memory performance. By enhancing memory capabilities, LLMs can improve in maintaining coherence, relevance, and human-like interaction over extended conversation. Improving these capabilities will help LLMs fulfill diverse user needs and expectations in various conversational scenarios.
Abstract (translated)
URL
https://arxiv.org/abs/2610.25463
PDF
-
To Manipulate or Not: Opinion Dynamics in LLM Interactions
2026-10-20 19:48:20Mohammad Siahkamari, Phuoc Dat Nguyen, Stephen MacNeilAbstract
Large language models increasingly take part in extended conversations where they can both shape and be shaped by another party's views, yet most persuasion research measures only one side of that exchange. We introduce a controlled LLM-LLM debate framework in which two language-model roles repeatedly discuss a contested topic (climate-change attribution): one is given an assigned stance and, in some conditions, an explicit instruction to persuade, while the other receives no assigned position. Across four ordered model combinations involving GPT-6 Luna and Gemini 2.5 Flash Lite, we track both each role's explicitly reported position and a separate measure of its turn-by-turn behavior inside the conversation. Opinion dynamics differ substantially across model combinations: same-model interactions narrow the opinion gap on average, clearly for only one of the two models, while cross-model interactions widen it. Explicit reports and in-dialogue behavior capture distinct aspects of opinion dynamics, and a blanket instruction to persuade does not reliably increase influence. These findings suggest that conversational influence depends not only on assigned roles and initial positions but also on the participating models, with implications for how opinion change is measured in extended LLM interactions.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28875
PDF
-
Small Average Error, Wide Individual Error: General-Purpose LLMs Estimating Energy Intake from Meal Photos in a Feeding Trial
Abstract
Photo-based estimation of food energy could ease dietary assessment, and general-purpose multimodal large language models (LLMs) make such estimates easy to obtain. Most evaluations use one photo taken before eating and compare with energy served, while dietary assessment needs energy eaten. We evaluated eight models from three vendors (Claude Haiku, Sonnet and Opus 5.5; GPT-6 luna, 6.1 sol and 6 astra; Gemini 3.8 Flash and 4 Argon) on 802 meals from ACETADA, a controlled feeding trial with weighed intake, with the before photo only, before and after photos in separate calls (subtraction), or both photos in one request. Three results ran against intuition. Subtraction under-estimated intake for six of eight models, because the models still report hundreds of kilocalories for a nearly empty tray. Scaling the before-only estimate by the average share eaten (0.87) matched much of the gain from the after photo; beyond that constant, the second photo lowered meal-level error by only 2.2 to 4.4 points, mainly in meals with food left. With one photo, two models had a near-zero intake bias only because they under-estimated the served energy and ignored uneaten food. In ablations, adding the after photo improved accuracy under both prompt formats for seven models, while prompt effects varied across models. Over 148 feeding days, five models kept the 90% confidence interval of their mean error within a pre-set 10% margin; one model also met the pre-set 30% limits of agreement, narrowly (-28% to +27%; limits outside 30% cannot be excluded), and the other seven did not. Protein was under-estimated by every model, most in protein-dense meals; repeated calls changed error by at most 1.4 points; a high reasoning effort helped only the smallest model. A low average bias is not evidence of accurate individual intake estimation; such estimates should support, not replace, assessment by a dietitian.
Abstract (translated)
URL
https://arxiv.org/abs/2610.26538
PDF
-
EHRAdapt: Adapting Pretrained Language Models to Electronic Health Records with Semantic Priors for Rare Clinical Events
2026-10-20 19:02:37Andre R Goncalves, Vincent Liu, Priyadip RayAbstract
Electronic health records (EHRs) encode clinical histories as (time, modality, code) tuples, whereas pretrained language models expect text tokens. Serializing tuples as text inflates sequence length and repeatedly encodes the same structure. We introduce EHRAdapt, an adapter that maps tuples directly into a frozen language model's embedding space. Modality and time gaps enter through a learned embedding and attention biases, and event codes receive dedicated vectors. Learning them is challenging because clinical vocabularies are long-tailed, leaving rare events with few observations. EHRAdapt therefore represents each event vector as the sum of a semantic prior and an evidence residual. The prior is a frozen, ontology-trained biomedical embedding of the event's clinical description, mapped into the input space by a shared learned projection, so it supplies clinical meaning even when observations are scarce. The residual is a learned low-rank, event-specific correction that refines it as evidence accumulates. We perform continued pretraining on records from about 4 million patients with three frozen LLM backbones, OLMo2 1B, Llama3.2 1B, and OLMo2 7B, training only the adapter (0.1--0.6% of all parameters). The full adapter outperforms all ablations in held-out next-event prediction on every backbone. Removing the semantic pathway hurts rare events most, over ten times more than the most frequent ones, whereas removing the residual hurts overall prediction but improves it for the rarest events. On reportable infectious-disease and syndromic classification, EHRAdapt outperforms text-based LLM and count-based baselines and the same LLMs adapted to the same records through next-event prediction, and both pathways improve rare-disease discrimination. The two pathways therefore play complementary roles, visible only when results are broken down by event frequency.
Abstract (translated)
URL
https://arxiv.org/abs/2609.34007
PDF
-
Decision language models as affordable and rapid alternatives to LLM judges of clinical generative AI
2026-10-20 18:46:47Rotem Gershon, Seffi Cohen, Dean Ariel, Havatzelet Bilavsky-Yarden, Amiel Sberro, Shai Yitzhaki, Osnat Tausky, Naama Golan, Maayan Rabinovitch, Ariel Avraham Hasidim, Itay Arianu, Maya Dagan, Ran Balicer, Noa DaganAbstract
Automated judges of clinical artificial intelligence (AI) outputs are necessary, but large language model (LLM) judges are costly and slow at scale. We compared Jev, a decision language model that returns structured judgments, with frontier and open-weight LLM judges in four use cases. On 118 pediatric answers, Jev's agreement with physicians' ordinal quality ratings (Gwet's AC2 0.50; expected score 0.53) matched or exceeded that of seven of 12 LLM judges, including frontier models, Claude Opus 5.5 (0.49) and GPT-6 (0.38). On physician-graded binary HealthBench criteria, Jev matched or exceeded four of six LLM judges. On MedHELM's open-ended tasks, Jev's agreement with a three-LLM jury approached or exceeded the jurors' internal agreement. On HealthBench Professional, Jev matched the benchmark's reference judge, GPT-5.4, on 91% of criteria (other judges, 91-93%). Frontier judges cost 73-294 times more than Jev and took 7-57 times longer. Jev may enable efficient clinical AI evaluation at scale.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28826
PDF
-
A Parallel Multi-Dialect Arabic Dependency Treebank
2026-10-20 18:45:40Noor Abo Mokh, Mohamed Adel, Jamila El-Gizuli, Go Inoue, Ahmed Elshabrawy, Nizar HabashAbstract
Dependency parsing resource development for Arabic has largely focused on Modern Standard Arabic, leaving dialectal Arabic comparatively under-resourced. We introduce MADARTB, the first large-scale parallel dependency treebank spanning five city dialects: Beirut, Cairo, Doha, Rabat and Tunis, and extending an existing parallel MSA version. In total MADARTB consists of 5,000 sentences annotated using the CATiB dependency scheme. We analyze syntactic similarity across all city dialects and MSA. Although global construction profiles exhibit high similarity across varieties, local syntactic structures frequently diverge, revealing substantial structural variation across dialects. We further establish a cross-dialect dependency parsing comparison using a supervised parser and four LLMs. The supervised parser achieves 87.5% LAS, while the best LLM setup trails at 85.7% LAS across all six varieties on the test set. We release our annotations and models publicly.
Abstract (translated)
URL
https://arxiv.org/abs/2610.28823
PDF
