Not All Tokens Are Created Equal: Query-Efficient Jailbreak Fuzzing for LLMs

Large Language Models(LLMs) are widely deployed, yet are vulnerable to jailbreak prompts that elicit policy-violating outputs. Although prior studies have uncovered these risks, they typically treat all tokens as equally important during prompt mutation, overlooking the varying contributions of individual tokens to triggering model refusals. Consequently, these attacks introduce substantial redundant searching under query-constrained scenarios, reducing attack efficiency and hindering comprehensive vulnerability assessment. In this work, we conduct a token-level analysis of refusal behavior and observe that token contributions are highly skewed rather than uniform. Moreover, we find strong cross-model consistency in refusal tendencies, enabling the use of a surrogate model to estimate token-level contributions to the target model's refusals. Motivated by these findings, we propose TriageFuzz, a token-aware jailbreak fuzzing framework that adapts the fuzz testing approach with a series of customized designs. TriageFuzz leverages a surrogate model to estimate the contribution of individual tokens to refusal behaviors, enabling the identification of sensitive regions within the prompt. Furthermore, it incorporates a refusal-guided evolutionary strategy that adaptively weights candidate prompts with a lightweight scorer to steer the evolution toward bypassing safety constraints. Extensive experiments on six open-source LLMs and three commercial APIs demonstrate that TriageFuzz achieves comparable attack success rates (ASR) with significantly reduced query costs. Notably, it attains a 90% ASR with over 70% fewer queries compared to baselines. Even under an extremely restrictive budget of 25 queries, TriageFuzz outperforms existing methods, improving ASR by 20-40%.

VORL-EXPLORE: A Hybrid Learning Planning Approach to Multi-Robot Exploration in Dynamic Environments

Hierarchical multi-robot exploration commonly decouples frontier allocation from local navigation, which can make the system brittle in dense and dynamic environments. Because the allocator lacks direct awareness of execution difficulty, robots may cluster at bottlenecks, trigger oscillatory replanning, and generate redundant coverage. We propose VORL-EXPLORE, a hybrid learning and planning framework that addresses this limitation through execution fidelity, a shared estimate of local navigability that couples task allocation with motion execution. This fidelity signal is incorporated into a fidelity-coupled Voronoi objective with inter-robot repulsion to reduce contention before it emerges. It also drives a risk-aware adaptive arbitration mechanism between global A* guidance and a reactive reinforcement learning policy, balancing long-range efficiency with safe interaction in confined spaces. The framework further supports online self-supervised recalibration of the fidelity model using pseudo-labels derived from recent progress and safety outcomes, enabling adaptation to non-stationary obstacles without manual risk tuning. We evaluate this capability separately in a dedicated severe-traffic ablation. Extensive experiments in randomized grids and a Gazebo factory scenario show high success rates, shorter path length, lower overlap, and robust collision avoidance. The source code will be made publicly available upon acceptance.

Omni-Manip: Beyond-FOV Large-Workspace Humanoid Manipulation with Omnidirectional 3D Perception

The deployment of humanoid robots for dexterous manipulation in unstructured environments remains challenging due to perceptual limitations that constrain the effective workspace. In scenarios where physical constraints prevent the robot from repositioning itself, maintaining omnidirectional awareness becomes far more critical than color or semantic information. While recent advances in visuomotor policy learning have improved manipulation capabilities, conventional RGB-D solutions suffer from narrow fields of view (FOV) and self-occlusion, requiring frequent base movements that introduce motion uncertainty and safety risks. Existing approaches to expanding perception, including active vision systems and third-view cameras, introduce mechanical complexity, calibration dependencies, and latency that hinder reliable real-time performance. In this work, We propose Omni-Manip, an end-to-end LiDAR-driven 3D visuomotor policy that enables robust manipulation in large workspaces. Our method processes panoramic point clouds through a Time-Aware Attention Pooling mechanism, efficiently encoding sparse 3D data while capturing temporal dependencies. This 360° perception allows the robot to interact with objects across wide areas without frequent repositioning. To support policy learning, we develop a whole-body teleoperation system for efficient data collection on full-body coordination. Extensive experiments in simulation and real-world environments show that Omni-Manip achieves robust performance in large-workspace and cluttered scenarios, outperforming baselines that rely on egocentric depth cameras.

MedCoRAG: Interpretable Hepatology Diagnosis via Hybrid Evidence Retrieval and Multispecialty Consensus

Diagnosing hepatic diseases accurately and interpretably is critical, yet it remains challenging in real-world clinical settings. Existing AI approaches for clinical diagnosis often lack transparency, structured reasoning, and deployability. Recent efforts have leveraged large language models (LLMs), retrieval-augmented generation (RAG), and multi-agent collaboration. However, these approaches typically retrieve evidence from a single source and fail to support iterative, role-specialized deliberation grounded in structured clinical data. To address this, we propose MedCoRAG (i.e., Medical Collaborative RAG), an end-to-end framework that generates diagnostic hypotheses from standardized abnormal findings and constructs a patient-specific evidence package by jointly retrieving and pruning UMLS knowledge graph paths and clinical guidelines. It then performs Multi-Agent Collaborative Reasoning: a Router Agent dynamically dispatches Specialist Agents based on case complexity; these agents iteratively reason over the evidence and trigger targeted re-retrievals when needed, while a Generalist Agent synthesizes all deliberations into a traceable consensus diagnosis that emulates multidisciplinary consultation. Experimental results on hepatic disease cases from MIMIC-IV show that MedCoRAG outperforms existing methods and closed-source models in both diagnostic performance and reasoning interpretability.

Learning to Draft: Adaptive Speculative Decoding with Reinforcement Learning

Speculative decoding accelerates large language model (LLM) inference by using a small draft model to generate candidate tokens for a larger target model to verify. The efficacy of this technique hinges on the trade-off between the time spent on drafting candidates and verifying them. However, current state-of-the-art methods rely on a static time allocation, while recent dynamic approaches optimize for proxy metrics like acceptance length, often neglecting the true time cost and treating the drafting and verification phases in isolation. To address these limitations, we introduce Learning to Draft (LTD), a novel method that directly optimizes for throughput of each draft-and-verify cycle. We formulate the problem as a reinforcement learning environment and train two co-adaptive policies to dynamically coordinate the draft and verification phases. This encourages the policies to adapt to each other and explicitly maximize decoding efficiency. We conducted extensive evaluations on five diverse LLMs and four distinct tasks. Our results show that LTD achieves speedup ratios ranging from 2.24x to 4.32x, outperforming the state-of-the-art method Eagle3 up to 36.4%.

An Optimization Method for Autoregressive Time Series Forecasting

Current time-series forecasting models are primarily based on transformer-style neural networks. These models achieve long-term forecasting mainly by scaling up the model size rather than through genuinely autoregressive (AR) rollout. From the perspective of large language model training, the traditional training process for time-series forecasting models ignores temporal causality. In this paper, we propose a novel training method for time-series forecasting that enforces two key properties: (1) AR prediction errors should increase with the forecasting horizon. Any violation of this principle is considered random guessing and is explicitly penalized in the loss function, and (2) the method enables models to concatenate short-term AR predictions for forming flexible long-term forecasts. Empirical results demonstrate that our method establishes a new state-of-the-art across multiple benchmarks, achieving an MSE reduction of more than 10% compared to iTransformer and other recent strong baselines. Furthermore, it enables short-horizon forecasting models to perform reliable long-term predictions at horizons over 7.5 times longer. Code is available at https://github.com/LizhengMathAi/AROpt

Boosting Maximum Entropy Reinforcement Learning via One-Step Flow Matching

Diffusion policies are expressive yet incur high inference latency. Flow Matching (FM) enables one-step generation, but integrating it into Maximum Entropy Reinforcement Learning (MaxEnt RL) is challenging: the optimal policy is an intractable energy-based distribution, and the efficient log-likelihood estimation required to balance exploration and exploitation suffers from severe discretization bias. We propose \textbf{F}low-based \textbf{L}og-likelihood-\textbf{A}ware \textbf{M}aximum \textbf{E}ntropy RL (\textbf{FLAME}), a principled framework that addresses these challenges. First, we derive a Q-Reweighted FM objective that bypasses partition function estimation via importance reweighting. Second, we design a decoupled entropy estimator that rigorously corrects bias, which enables efficient exploration and brings the policy closer to the optimal MaxEnt policy. Third, we integrate the MeanFlow formulation to achieve expressive and efficient one-step control. Empirical results on MuJoCo show that FLAME outperforms Gaussian baselines and matches multi-step diffusion policies with significantly lower inference cost. Code is available at https://github.com/lzqw/FLAME.

TriSpec: Ternary Speculative Decoding via Lightweight Proxy Verification

Inference efficiency in Large Language Models (LLMs) is fundamentally limited by their serial, autoregressive generation, especially as reasoning becomes a key capability and response sequences grow longer. Speculative decoding (SD) offers a powerful solution, providing significant speed-ups through its lightweight drafting and parallel verification mechanism. While existing work has nearly saturated improvements in draft effectiveness and efficiency, this paper advances SD from a new yet critical perspective: the verification cost. We propose TriSpec, a novel ternary SD framework that, at its core, introduces a lightweight proxy to significantly reduce computational cost by approving easily verifiable draft sequences and engaging the full target model only when encountering uncertain tokens. TriSpec can be integrated with state-of-the-art SD methods like EAGLE-3 to further reduce verification costs, achieving greater acceleration. Extensive experiments on the Qwen3 and DeepSeek-R1-Distill-Qwen/LLaMA families show that TriSpec achieves up to 35\% speedup over standard SD, with up to 50\% fewer target model invocations while maintaining comparable accuracy.

HeaPA: Difficulty-Aware Heap Sampling and On-Policy Query Augmentation for LLM Reinforcement Learning

RLVR is now a standard way to train LLMs on reasoning tasks with verifiable outcomes, but when rollout generation dominates the cost, efficiency depends heavily on which prompts you sample and when. In practice, prompt pools are often static or only loosely tied to the model's learning progress, so uniform sampling can't keep up with the shifting capability frontier and ends up wasting rollouts on prompts that are already solved or still out of reach. Existing approaches improve efficiency through filtering, curricula, adaptive rollout allocation, or teacher guidance, but they typically assume a fixed pool-which makes it hard to support stable on-policy pool growth-or they add extra teacher cost and latency. We introduce HeaPA (Heap Sampling and On-Policy Query Augmentation), which maintains a bounded, evolving pool, tracks the frontier using heap-based boundary sampling, expands the pool via on-policy augmentation with lightweight asynchronous validation, and stabilizes correlated queries through topology-aware re-estimation of pool statistics and controlled reinsertion. Across two training corpora, two training recipes, and seven benchmarks, HeaPA consistently improves accuracy and reaches target performance with fewer computations while keeping wall-clock time comparable. Our analyses suggest these gains come from frontier-focused sampling and on-policy pool growth, with the benefits becoming larger as model scale increases. Our code is available at https://github.com/horizon-rl/HeaPA.

TeachBench: A Syllabus-Grounded Framework for Evaluating Teaching Ability in Large Language Models

Large language models (LLMs) show promise as teaching assistants, yet their teaching capability remains insufficiently evaluated. Existing benchmarks mainly focus on problem-solving or problem-level guidance, leaving knowledge-centered teaching underexplored. We propose a syllabus-grounded evaluation framework that measures LLM teaching capability via student performance improvement after multi-turn instruction. By restricting teacher agents to structured knowledge points and example problems, the framework avoids information leakage and enables reuse of existing benchmarks. We instantiate the framework on Gaokao data across multiple subjects. Experiments reveal substantial variation in teaching effectiveness across models and domains: some models perform well in mathematics, while teaching remains challenging in physics and chemistry. We also find that incorporating example problems does not necessarily improve teaching, as models often shift toward example-specific error correction. Overall, our results highlight teaching ability as a distinct and measurable dimension of LLM behavior.