NeurIPS 2026 Papers

Optimization under heavy-tailed stochastic noise remains a major challenge in modern machine learning, where classical moment assumptions often fail to capture realistic gradient distributions. In this work, we revisit robust estimation techniques and provide a systematic analysis of the Median-of-Means (MoM) estimator under a structured noise model that interpolates between a symmetric heavy-tailed distribution with bounded $\beta$-th moment ($\beta \in (0, 1]$) and a mean-zero distribution with bounded $\alpha$-th moment ($\alpha \in (1, 2]$). We derive new upper bounds on the variance and bias of MoM in this setting, thereby extending the applicability of existing results beyond purely symmetric or bounded-moment settings. Building on these properties, we study clipped stochastic optimization methods with MoM gradient estimation and establish high-probability convergence guarantees that hold for general $\beta \in (0, 1]$ and $\alpha \in (1, 2]$. Our bounds match or improve upon prior complexity guarantees while requiring weaker distributional assumptions, and they reveal that bias only affects the final neighborhood of convergence. Overall, this work advances the theoretical foundations for robust stochastic optimization in the presence of extreme gradient noise.

Large language models (LLMs) have shown promise in medical question answering and clinical reasoning, yet their improvement remains constrained by static parametric knowledge and costly expert supervision. Self-evolving agents offer a promising alternative by enabling models to improve through iterative task generation and problem-solving. However, most existing self-evolving methods are designed for easily verifiable domains such as mathematics and coding, where solutions can be checked by exact answers or executable programs. Medical reasoning is fundamentally different: it is open-ended, knowledge-intensive, and often only partially verifiable. We present \textbf{MedZERO}, a self-evolving framework for open-ended medical reasoning. MedZERO couples an \textit{Examiner} that generates frontier medical question-option pairs with a \textit{Reasoner} that solves them through evidence-grounded multi-turn reasoning with external knowledge tools. To support reliable, continual improvement, MedZERO introduces controlled knowledge accumulation, which maintains temporary exploratory knowledge and curated persistent knowledge in reasoning. We evaluate MedZERO on five public medical reasoning benchmarks using 4B- and 8B-scale base models under open-ended evaluation. Across all settings, MedZERO consistently outperforms the underlying base models and prior self-evolving baselines, achieving up to 13.7 average accuracy-point gains over the next-best self-evolving baseline.

Multimodal medical AI requires a common visual interface that integrates heterogeneous imaging evidence for patient condition modeling. However, existing medical visual tokenizers are often tied to a specific dimension and imaging modality, forcing multimodal systems to rely on fragmented and poorly reusable representations. We introduce MedVTok, a general-purpose visual tokenizer that maps 2D images and 3D volumes from diverse imaging modalities into a unified token space. Building such a tokenizer poses two key challenges: preserving anatomical structure across dimensions and capturing clinical semantics across modalities. To this end, MedVTok introduces (i) slice differential regularization, which explicitly models inter-slice anatomical coherence missing in previous per-pixel/per-voxel supervision, and (ii) multi-expert representation alignment, which integrates knowledge from multiple medical expert encoders while retaining modality-specific complementary knowledge. Trained on 28M images and 70K volumes across 9 imaging modalities, MedVTok supports classification, retrieval, segmentation, synthesis, visual question answering, and medical report generation through a single tokenizer. Across 24 benchmarks, MedVTok achieves state-of-the-art performance, demonstrating a scalable token interface that unifies medical image dimension, modality, and task usage. Code and weights will be released.

Medical image interpretation requires diagnoses grounded in case-specific visual evidence. However, medical vision-language models often produce plausible answers by exploiting clinical language priors and report-level co-occurrence patterns rather than faithfully using the input image, leading to Visual De-anchoring and Spatial-Semantic Binding Breakdown, where reasoning detaches from visual evidence or binds correct answers to incorrect anatomical support. In this work, we introduce MedVIGOR, a framework that internalizes visual evidence as an intrinsic constraint on medical VLM reasoning. MedVIGOR combines visual-necessity supervision, dual-pathway visual anchor internalization with MedSAM-derived spatial support and DINOv2-derived semantic patterns, and evidence-consistent reinforcement learning to preserve alignment among visual evidence, intermediate reasoning, and final answers. We further present MedVIGOR-Bench, a unified benchmark for evaluating whether medical VLMs are both diagnostically correct and evidentially grounded across concept recognition and visual localization tasks. Experiments show that MedVIGOR improves diagnostic accuracy and grounding fidelity over strong medical VLM baselines, highlighting the value of internalized visual evidence for trustworthy medical reasoning.

Medical language models promise expert clinical support, but the strongest open source LLMs remain too large for private, edge devices deployment, while compact medical models still trail substantially on knowledge-intensive and real-world healthcare tasks. We present MedPsy, a family of text-only 1.7B and 4B medical language models designed for edge deployment. Our recipe combines a synthetic medical data pipeline over biology, medicine, and health seeds with chain-of-thought targets from a 235B medical-focused LLM teacher; a four-stage post-training curriculum of two SFT stages followed by two RL stages with hard-sample mining; and a mobile-oriented quantization study over several GGUF variants per model. On seven closed-ended medical benchmarks, MedPsy-4B surpasses MedGemma-1.5-4B by +19.34 points and matches MedGemma-27B while being 6.75x smaller; MedPsy-1.7B outperforms MedGemma-1.5-4B by +11.42 points despite being less than half its size. On HealthBench-Hard, MedPsy-4B surpasses MedGemma-27B by +15.33 and MedPsy-1.7B surpasses it by +11.66 points despite being 16x smaller. Beyond accuracy, MedPsy reduces average response length by 1.7x (1.7B) and 3.2x (4B) versus its Qwen3 backbones, and 4-bit quantization retains accuracy within 1 point of BF16 while reducing disk footprint by ~69%, enabling practical deployment on resource-constrained edge devices.

Large language models (LLMs) now reach expert-level scores on medical licensing exams, encouraging the assumption that high scores imply safe medical judgment while patients increasingly use them for health advice. We show this assumption is fragile: when misleading context is injected into questions that LLMs originally answer correctly, they abandon the correct answer. We call the ability to maintain correct judgment under adversarial context epistemic resilience, and introduce MedMisBench to measure it. MedMisBench contains 10,932 medical question items and 48,889 misleading context-option pairs spanning medical reasoning, agentic capability, and patient-journey evaluation. Across 11 model configurations, mean accuracy falls from 71.1% on original questions to 38.0% under focused misleading context, with 51.5% attack success. The most damaging injections are formal, rule-like fabrications: authority-framed falsehoods reach 69.5% attack success and exception-poisoning claims reach 64.1%. A 14-member clinical panel from 7 countries identified serious potential harm in 38.2% of reviewed cases. MedMisBench exposes a structural blind spot in LLM evaluation in medical settings: existing benchmarks measure what models know, but not whether they preserve correct medical judgment under misleading context.

Constantly evolving real-world knowledge necessitates models to be updated continuously. Especially in medicine, as clinical evidence changes over time, outdated knowledge can pose safety risks. Existing evaluations of knowledge integration focus on factual recall, offering limited insight into whether newly integrated knowledge is actually usable. Our benchmark __MedKIT__ (Medical Knowledge Integration and Transfer) provides a granular evaluation of how models integrate and apply knowledge under realistic sequences of clinical updates. Each instance corresponds to a factual update derived from clinical evidence, paired with targeted probes that assess transfer across lexical variation, relational transformations, compositional reasoning, and open-ended operationalization, as well as locality tests for knowledge preservation. Using MedKIT, we conduct a large-scale empirical study of 12 knowledge integration strategies across 5 diverse models, including both general-purpose and medical LLMs. Our results reveal a consistent gap between recall and usable knowledge: while most methods achieve strong gains on the original update task and under lexical variation, relational generalization is limited, and no method yields meaningful improvements on compositional or operational tasks. These findings highlight a fundamental challenge in knowledge integration and position MedKIT as a testbed for developing methods that make newly integrated knowledge more consistently usable across tasks and contexts.

Medical illustration generation is not merely domain-specific text-to-image (T2I) generation, but a high-constraint visual reasoning problem where visually plausible images can be invalid due to anatomical and structural errors. Existing T2I models largely rely on one-pass generation and lack an internal mechanism to inspect and correct such biomedical inconsistencies. We introduce **MedIGen**, a unified medical illustration generation model that transforms one-pass synthesis into an **interleaved introspection-aware refinement** process, coupling generation with self-reflection and re-generation through “generate–reflect–refine” cycles. MedIGen is trained with three progressive stages: (1) large-scale medical illustration pretraining on 1.21M samples, (2) mixture training over generation, reflection, and refinement sub-skills, and (3) reinforcement learning with dual-level rewards that optimize both reasoning validity and final visual correctness. We further present **IlluGenBench**, an expert-aligned, rubric-driven benchmark with 296 diverse tasks and 9,015 criteria, evaluating scientific accuracy, structural correctness, and semantic alignment beyond coarse visual plausibility. Experiments show that MedIGen substantially outperforms strong open-source unified and reasoning-based baselines, establishing a new open-source frontier. All resources will be open-sourced to facilitate future research.

Medical multimodal large language models (MLLMs) have advanced image understanding and short-video analysis, but real clinical review often requires full-procedure video understanding. Unlike general long videos, medical procedures contain highly redundant anatomical views, while decisive evidence is temporally sparse, spatially subtle, and context dependent. Existing benchmarks often assume this evidence has already been localized through images, short clips, or pre-segmented videos, leaving the retrieval-before-reasoning problem under-tested. We introduce MedHorizon, an in-the-wild benchmark for long-context medical video understanding. MedHorizon preserves 759 hours of full-length clinical procedures and provides 1,253 evidence-grounded multiple-choice questions that jointly evaluate sparse evidence understanding and multi-hop clinical reasoning. Its evidence is extremely sparse, with only 0.166% evidence frames on average, requiring models to search noisy procedural streams before interpreting and aggregating findings. We evaluate representative general-domain, medical-domain, and long-video MLLMs. The best model reaches only 41.1% accuracy, showing that current systems remain far from robust full-procedure understanding. Further analysis yields four key findings: performance does not scale reliably with more frames, evidence retrieval and clinical interpretation remain primary bottlenecks; these bottlenecks are rooted in weak procedural reasoning and attention drift under redundancy, and generic sampling methods only partially balances local detail with global coverage. MedHorizon provides a rigorous testbed for MLLMs that retrieve sparse evidence and reason over complete clinical workflows.

Clinical evidence is inherently relational: medical images are interpreted with prior studies, cross-modal examinations, reports, and localized findings, and retrieving relevant evidence is therefore central to clinical interpretation. Recent foundation models increasingly encode these heterogeneous sources into shared embedding spaces, yet existing embedding evaluation often focuses on isolated matches, leaving unclear whether current models retrieve relevant evidence across changes in modality, appearance, and granularity. We introduce MedHEB, a unified benchmark with $107$ retrieval tasks from $54$ datasets, including both in-distribution and held-out-source evaluation, across three axes: evidence form (2D images, 3D volumes, text), pairing type (image--image, image--text, text--text, and modality-bridging visual matching), and relevance granularity (global, question-conditioned, region-level). Under a single ranking protocol, MedHEB varies clinically relevant evidence definitions to test whether embedding proximity remains meaningful across heterogeneous evidence, including modality-bridging visual retrieval and region-level text-to-region retrieval. Across representative models, performance is uneven across retrieval families: models strong on conventional image--text or text--text retrieval are less reliable on modality-bridging and region-level tasks. Current embedding spaces capture some clinical relationships well but do not yet provide balanced retrieval across heterogeneous evidence. We further provide MedEmb, an MLLM-based reference embedding model trained on the MedHEB ID split, as a reproducible baseline for unified 2D--3D--text medical retrieval. We release MedHEB with preprocessing scripts, evaluation code, access manifests, and MedEmb to support research on clinically meaningful medical embeddings.

Medical imaging benchmarks often evaluate VLMs on pre-selected 2D images, slices, crops, or patches, making evaluation closer to visual recognition. Real clinical workflows impose a different burden: readers must search through complete studies, operate imaging software, navigate across slices and magnifications, and document visual evidence that can be audited. We argue that this evidence-producing workflow is a critical missing evaluation axis for medical imaging agents. To study it, we introduce MedFlowBench, a full-study benchmark for VLM agents, together with MedOpenClaw, a controlled and replayable runtime in which agents operate medical imaging viewers such as 3D Slicer and QuPath. In each episode, an agent inspects a complete radiology study or whole-slide pathology image, returns a task answer, and submits structured evidence, including key slices, coordinates, regions of interest, or lesion-state fields. This evidence is automatically checked against withheld masks, annotations, and labels. Across evaluated models, final answer-only scoring gives an overly optimistic picture: when answers must also be supported by correct evidence, performance drops substantially on complex workflows. We further find that adding image-analysis tools does not by itself solve the problem. Tools help when they make a complex procedure simple and reliable, but agents still struggle when they must choose inputs, manage viewer state, and verify intermediate outputs over multiple steps. MedFlowBench exposes whether medical imaging agents can produce auditable evidence from complete studies, rather than plausible answers from selected images.

Real-world clinical diagnosis is a complex process in which the doctor is required to obtain information from both interaction with the patient and conducting medical exams. Additionally, the doctor needs to adapt to different patient personas, as well as noisy and incomplete information that can happen at any time during the process. However, existing benchmarks for medical LLMs and methods for automatic diagnosis largely simplify this process by reducing it to single-turn question answering, noise-free conversations, or sequential exam making, etc., ignoring the interactive and uncertain nature of clinical diagnosis. In this paper, we aim to address this gap by formalizing clinical diagnosis as a Partially Observable Markov Decision Process (POMDP) with three action types: questioning the patient, ordering medical exams as tool calls, and issuing a diagnosis. We also introduce a systematic noise model comprising seven patient noise types and three exam noise types. Using our proposed environment, we train an effective diagnosis agent, \textbf{MedExAgent}, through a two-stage pipeline that first performs supervised finetuning on synthetic conversations structured after the Calgary-Cambridge model for clinical interviews, and then applies DAPO to optimize a composite reward capturing diagnostic accuracy, tool call quality, and exam cost including financial cost and patient discomfort. Through extensive experiments and ablation studies, we demonstrate that MedExAgent achieves diagnostic performance comparable to larger models while maintaining cost-efficient examination strategies.

Doctor agents are moving beyond single-turn answer generation toward evolving clinical decision systems that acquire evidence within episodes and reuse experience across episodes. Fixed-input medical QA and many interactive benchmarks provide limited support for evaluating how episode-level decisions interact with cross-episode experience. We introduce MedEvoEval, an executable longitudinal evaluation framework based on action-gated simulated outpatient episodes. Each case is converted into role-specific patient, examination, and manager views; evidence is revealed through valid actions, and structured traces link observations, actions, final outputs, scores, and optional experience write-back. Experiments show that these traces expose process costs hidden by final-answer scoring, characterize how MDT-style consultation reallocates resources, and support analyses of memory maturation, held-out transfer, update-stage response, and backward retention. On 200 held-out cases, the mature-memory condition with sequential write-back reduces interaction turns by 0.915 and improves composite utility by 0.0483 after Holm correction. MedEvoEval provides a concrete basis for evaluating experience-driven improvement, transfer, and retention. The accompanying artifact provides 700 processed episode records with provenance, role-specific schemas, and execution, scoring, and analysis code.

Video world models are now being adapted to medical applications such as surgical simulation, procedural rehearsal, and synthetic data generation, but their inference cost remains a major obstacle to practical use. Training-free caching is attractive because it can be applied to existing checkpoints, yet current methods face a clear trade-off. Lightweight rules track only the average change in the video and treat all regions of the frame as equally important, which is a poor fit for medical video where the meaningful motion is often confined to a small region such as a tool tissue interaction. Richer rules recover spatial awareness by partially running the network on every step to inform the skip decision, but this decision cost is paid whether the step is skipped or not, and it eats into the speedup. We introduce \textbf{MedCache}, a training-free and probe-free cache for rectified-flow video generation. MedCache reads a spatial risk signal directly from the latent using cheap arithmetic, with no help from the network itself. It combines this risk signal with a prompt-derived domain prior and a lazy consistency check that runs only when reuse is no longer clearly safe. Across three open-world model families in Text2World and Image2World settings, MedCache improves both speed and quality over the strongest training free baselines in the medical regime it targets. On Cosmos-H-Surgical 2B Text2World, it raises Total quality from 0.767 to 0.837 over EasyCache while reducing latency from 185 seconds to 95 seconds. Ablations show that risk-weighted drift drives the runtime gain, while the prior and guard recover fidelity.

Large language models (LLMs) in expert domains face a trade-off between \emph{agentic tool use}, which externalizes expertise behind brittle multi-turn runtimes, and \emph{end-to-end specialization}, which internalizes expertise only implicitly at task-level granularity. We propose \emph{internalized meta-capabilities}: atomic, parametrically internalized, and composable reasoning units that an LLM can select and combine within a single chain-of-thought (CoT). We instantiate this idea as \textbf{Med-Agentic} for medical visual diagnosis, where the model emits a capability-tagged structured CoT in one forward pass, turning capability invocation into tagged token generation rather than external tool calls. For training, \textbf{Med-OPD} performs multi-teacher on-policy distillation with \emph{token-level dual-factor teacher routing} conditioned on both the image and the active capability tag, strictly generalizing per-prompt teacher routing. We show that this routing gives an unbiased estimator of the same expected teacher-KL gradient. Across public medical imaging datasets spanning nine anatomies and eight modalities, Med-Agentic consistently improves in-distribution and out-of-distribution diagnostic accuracy over the strong competitors.

NPU design-space exploration for LLM inference mainly relies on an expensive simulator that often returns only scalar feedback such as latency and power. Such feedback can rank evaluated configurations, but gives little information for deciding why a configuration is slow or which architecture parameter should be changed next. This limitation is especially severe because prefill and decode workloads can be limited by different resources, including compute throughput, HBM bandwidth, and inter-chip communication. We propose CriticNPU, a sample-efficient DSE framework built around an online-calibrated mechanistic critic for NPU architecture design. The critic uses explicit hardware limits to evaluate NPU configurations: before simulation, it estimates whether a proposed configuration is likely to improve over the current best and filters weak candidates; after simulation, it aggregates operator-level runtimes into prefill/decode bottleneck summaries and predicts the effect of candidate parameter changes. We instantiate the critic with a Roofline-style model that derives compute, memory, and communication limits from architecture parameters and calibrates per-operator-type effective limits using simulator observations. And CriticNPU combines this critic with a workload extractor and an architecture planner to propose configurations. Across six LLM workloads and 16 deployment configurations per model, CriticNPU achieves best-or-tied-best latency in 74% of prefill and 87.5% of decode settings while using roughly half the simulator calls of competitive DSE baselines. We conduct extensive ablation studies and detailed analyses to validate the effectiveness of our method from multiple perspectives.

The integration of AI Overviews into search engines enhances user experience but diverts traffic from content creators, potentially discouraging high-quality content creation and causing user attrition that undermines long-term search engine profit. To address this issue, we propose a game-theoretic model of creator competition with costly effort, characterize equilibrium behavior, and design two incentive mechanisms: a \emph{citation mechanism} that references sources within an AI Overview, and a \emph{compensation mechanism} that offers monetary rewards to creators. For both cases, we provide structural insights for profit-maximizing mechanisms. Evaluations parameterized by real click data show that although AI Overviews harm long-term search engine profit, interventions based on our proposed mechanisms can increase long-term profit across a range of realistic scenarios, pointing toward a more sustainable trajectory for AI-enhanced search ecosystems.

Undesired model behaviors can emerge or re-emerge at any stage of training. Once observed, how effectively can we suppress them? To study this question, we first construct two model organisms---language models trained to exhibit a specific misalignment---with qualitatively distinct structures. In the narrow-trigger organism, undesired behavior is elicited by an identifiable prompt feature. In the broad-trigger organism, misalignment surfaces diffusely across open-ended responses without a separable cue. We then formalize post-hoc behavioral suppression and evaluate methods along three criteria: immediate suppression of the target behavior, durability under further training, and preservation of general capability. We compare standard methods---SFT, GRPO, and gradient-ascent unlearning---along these dimensions while controlling the number of gradient updates. Building on prior work on alignment shallowness, we hypothesize that more durable suppression requires the ability to recover from a wider range of misalignment states. We introduce Prefix-Expanded Adversarial Reinforcement Learning (PEARL), a GRPO variant that augments rollouts with continuations from intermediate states sampled along cached organism trajectories. Across both organism settings and two model families (Qwen3-4B, GPT-OSS-20B), PEARL achieves lower exploit rates while preserving task accuracy, and yields lower reactivation rates under both targeted and benign capability fine-tuning than SFT and GRPO baselines. Together, our framework and method offer a step toward more durable post-hoc removal of undesired behaviors.

As user simulators are increasingly used for interactive training and evaluation of AI assistants, it is essential that they represent the diverse behaviors of real users. While existing works train user simulators to generate human-like responses, whether they capture the broad and heterogeneous distribution of real user behaviors remains an open question. In this work, we introduce a method to measure the distributional gap between real and simulated user behaviors, validated through a human study and ablations. Given a dataset of real and simulated conversations, our method extracts representations of user behavior from each conversation, quantizes them into discrete distributions via clustering, then computes divergence metrics. We provide the first systematic evaluation of 24 LLM-based user simulators on coding and writing tasks, and reveal a large distributional gap from real users that varies across model families, scales, and behavioral facets. Pairwise comparisons show that most simulators behave similarly, while a few stand apart. Combining behaviorally complementary simulators brings the resulting distribution closer to real users compared to either simulator on its own. Finally, a TF-IDF analysis of the clusters surfaces interpretable patterns of behaviors that simulators capture, miss, and hallucinate.