Large Multimodal Models (LMMs) have achieved remarkable success in visual understanding, yet they struggle with knowledge-intensive queries involving long-tail entities or evolving information due to static parametric knowledge. Recent search-augmented approaches attempt to address this limitation, but existing methods rely on indiscriminate whole-image retrieval that introduces substantial visual redundancy and noise, and lack deep iterative reflection, limiting their effectiveness on complex visual queries. To overcome these challenges, we propose Glance-or-Gaze (GoG), a fully autonomous framework that shifts from passive perception to active visual planning. GoG introduces a Selective Gaze mechanism that dynamically chooses whether to glance at global context or gaze into high-value regions, filtering irrelevant information before retrieval. We design a dual-stage training strategy: Reflective GoG Behavior Alignment via supervised fine-tuning instills the fundamental GoG paradigm, while Complexity-Adaptive Reinforcement Learning further enhances the model’s capability to handle complex queries through iterative reasoning. Experiments across six benchmarks demonstrate state-of-the-art performance. Ablation studies confirm that both Selective Gaze and complexity-aware RL are essential for effective visual search. We will release our data and models for further exploration soon.
论文检索
输入标题、作者或关键词，从 1,248 篇学术成果中精准定位
ProCeedRL: Process Critic with Explorative Demonstration Reinforcement Learning for LLM Agentic Reasoning
PDF ↗Reinforcement Learning (RL) significantly enhances the reasoning abilities of large language models (LLMs), yet applying it to multi-turn agentic tasks remains challenging due to the long-horizon nature of interactions and the stochasticity of environmental feedback.We identify a structural failure mode in agentic exploration: suboptimal actions elicit noisy observations into misleading contexts, which further weaken subsequent decision-making, making recovery increasingly difficult.This cumulative feedback loop of errors renders standard exploration strategies ineffective and susceptible to the model’s reasoning and the environment’s randomness.To mitigate this issue, we propose **ProCeedRL**: **Pro**cess **C**ritic with **E**xplorativ**e** **D**emonstration RL, shifting exploration from passive selection to active intervention.ProCeedRL employs a process-level critic to monitor interactions in real time, incorporating reflection-based demonstrations to guide agents in stopping the accumulation of errors.We find that this approach significantly exceeds the model’s saturated exploration performance, demonstrating substantial exploratory benefits.By learning from exploratory demonstrations and on-policy samples, ProCeedRL significantly improves exploration efficiency and achieves superior performance on complex deep search and embodied tasks.
Despite the rapid advancements in Multimodal Large Language Models (MLLMs), a critical question regarding their visual grounding mechanism remains unanswered: do these models genuinely read text embedded in images, or do they merely rely on parametric shortcuts in the text prompt? In this work, we diagnose this issue by introducing the Visualized-Question (VQ) setting, where text queries are rendered directly onto images to structurally mandate visual engagement. Our diagnostic experiments on Qwen2.5-VL reveal a startling capability-utilization gap: despite possessing strong OCR capabilities, models suffer a performance degradation of up to 12.7% in the VQ setting, exposing a deep-seated modality laziness. To bridge this gap, we propose SimpleOCR, a plug-and-play training strategy that imposes a structural constraint on the learning process. By transforming training samples into the VQ format with randomized styles, SimpleOCR effectively invalidates text-based shortcuts, compelling the model to activate and optimize its visual text extraction pathways. Empirically, SimpleOCR yields robust gains without architectural modifications. On four representative OOD benchmarks, it surpasses the base model by 5.4% and GRPO based on original images by 2.7%, while exhibiting extreme data efficiency, achieving superior performance with 30x fewer samples (8.5K) than recent RL-based methods. Furthermore, its plug-and-play nature allows seamless integration with advanced RL strategies like NoisyRollout to yield complementary improvements. Code is available at https://github.com/aiming-lab/SimpleOCR.
Large Language Models (LLMs) have become a popular interface for human–AI interaction, supporting information seeking and task assistance through natural, multi-turn dialogue. To respond to users within multi-turn dialogues, the context-dependent user intent evolves across interactions, requiring contextual interpretation, query reformulation, and dynamic coordination between retrieval and generation. Existing studies usually follow static “rewrite, retrieve, and generate” pipelines, which optimize different procedures separately and overlook the mixed-initiative action optimization simultaneously. Although the recent developments in deep search agents demonstrate the effectiveness in jointly optimizing retrieval and generation via reasoning, these approaches focus on single-turn scenarios, which might lack the ability to handle multi-turn interactions. We introduce a conversational agent that interleaves search and reasoning across turns, enabling exploratory and adaptive behaviors learned through reinforcement learning (RL) training with tailored rewards towards evolving user goals. The experimental results across four widely used conversational benchmarks demonstrate the effectiveness of our methods by surpassing several existing strong baselines.
Large language models (LLMs) augmented with multi-step reasoning and action generation abilities have shown promise in leveraging external tools to tackle complex tasks that require long-horizon planning. However, existing approaches either rely on implicit planning in the reasoning stage or introduce explicit planners without systematically addressing how to optimize the planning stage. As evidence, we observe that under vanilla reinforcement learning (RL), planning tokens exhibit significantly higher entropy than other action tokens, revealing uncertain decision points that remain under-optimized. To address this, we introduce DeepPlanner, an end-to-end RL framework that effectively enhances the planning capabilities of deep research agents. Our approach shapes token-level advantage with an entropy-based term to allocate larger updates to high entropy tokens, and selectively upweights sample-level advantages for planning-intensive rollouts. Extensive experiments across seven deep research benchmarks demonstrate that DeepPlanner improves planning quality and achieves state-of-the-art results under a substantially lower training budget.
Large language models (LLMs) are post-trained through reinforcement learning (RL) to evolve into Reasoning Language Models (RLMs), where the hallmark of this advanced reasoning is “aha” moments when they start to perform strategies, such as self-reflection and deep thinking, within chain of thoughts (CoTs). Motivated by this, this paper proposes a novel reinforced strategy injection mechanism (rSIM), that enables any LLM to become an RLM by employing a small planner to guide the LLM’s CoT through the adaptive injection of reasoning strategies. To achieve this, the planner (leader agent) is jointly trained with an LLM (follower agent) using multi-agent RL (MARL), based on a leader-follower framework and straightforward rule-based rewards. Experimental results show that rSIM enables Qwen2.5-0.5B to become an RLM and significantly outperform Qwen2.5-14B across mathematical, coding, and financial reasoning tasks. Moreover, the planner is generalizable: it only needs to be trained once and can be applied as a plug-in to substantially improve the reasoning capabilities of existing LLMs. In addition, the planner supports continual learning across various tasks, allowing its planning abilities to gradually improve and generalize to a wider range of problems. Our source code is available under the examples/rSIM of https://github.com/AgenticFinLab/eparl.
Large language model (LLM)-based agents are increasingly deployed in e-commerce shopping. To perform thorough, user-tailored product searches, agents should interpret personal preferences, engage in multi-turn dialogues, and ultimately retrieve and discriminate among highly similar products. However, existing research has yet to provide a unified simulation environment that consistently captures all of these aspects, and always focuses solely on evaluation benchmarks without training support. In this paper, we introduce ShopSimulator, a large-scale and challenging Chinese shopping environment. Leveraging ShopSimulator, we evaluate LLMs across diverse scenarios, finding that even the best-performing models achieve less than 40% full-success rate. Error analysis reveals that agents struggle with deep search and product selection in long trajectories, fail to balance the use of personalization cues, and to effectively engage with users. Further training exploration provides practical guidance for overcoming these weaknesses, with the combination of supervised fine-tuning (SFT) and reinforcement learning (RL) yielding significant performance improvements.
Chaining the Evidence: Robust Reinforcement Learning for Deep Search Agents with Citation-Aware Rubric Rewards
PDF ↗Reinforcement learning (RL) has emerged as a critical technique for enhancing LLM-based deep search agents. However, existing approaches primarily rely on binary outcome rewards, which fail to capture the comprehensiveness and factuality of agents’ reasoning process, and often lead to undesirable behaviors such as shortcut exploitation and hallucinations. To address these limitations, we propose Citation-aware Rubric Rewards (CaRR), a fine-grained reward framework for deep search agents that emphasizes reasoning comprehensiveness, factual grounding, and evidence connectivity. CaRR decomposes complex questions into verifiable single-hop rubrics and requires agents to satisfy these rubrics by explicitly identifying hidden entities, supporting them with correct citations, and constructing complete evidence chains that link to the predicted answer. We further introduce Citation-aware Group Relative Policy Optimization (C-GRPO), which combines CaRR and outcome rewards for training robust deep search agents. Experiments show that C-GRPO consistently outperforms standard outcome-based RL baselines across multiple deep search benchmarks. Our analysis also validates that C-GRPO effectively discourages shortcut exploitation, promotes comprehensive, evidence-grounded reasoning, and exhibits strong generalization to open-ended deep research tasks.
Deep Learning · Large Language Models
Reinforcement learning (RL), particularly RL from verifiable reward (RLVR), has become a crucial phase of training large language models (LLMs) and a key focus of current scaling efforts. However, optimization practices in RL largely follow those of next-token-prediction stages (e.g., pretraining and supervised fine-tuning), despite the fundamental differences between RL and these stages emphasized by recent work. One such practice is the use of the AdamW optimizer, which is widely adopted for training large-scale transformers despite its high memory overhead. Our analysis shows that both momentum and adaptive learning rate of AdamW are less influential in RL than in SFT, leading us to hypothesize that RL benefits less from Adam’s per-parameter adaptive learning rates and momentum. Confirming our hypothesis, our experiments demonstrate that the substantially more memory-efficient SGD, which is known to perform poorly in supervised learning of large-scale transformers, matches or even outperforms AdamW in RL for LLMs. Remarkably, full fine-tuning with SGD updates fewer than 0.02% of model without any sparsity-promoting regularization, more than 1,000 times fewer than AdamW. Our analysis offers potential reasons for this update sparsity. Our findings provide fresh insights into the optimization dynamics of RL in LLMs and demonstrate that RL can be substantially more parameter-efficient than previously recognized.
Deep Learning · Large Language Models
Online reinforcement learning is widely used to align large language models (LLMs) with reward signals, yet training can be unstable under noisy or misspecified rewards. We identify a failure mode we call directional inconsistency: within a batch, a small set of high-reward rollouts induces representation-space preference directions that sharply disagree with the batch majority, resulting in high-variance and destabilizing updates. We propose GeoAlign, a lightweight plug-in for rollout curation in iterative policy optimization. GeoAlign (i) forms within-prompt preference pairs, (ii) learns an online projector on per-rollout hidden states to concentrate reward-ordered displacement directions, and (iii) detects directionally inconsistent rollouts via their angular deviation from a batch consensus prototype and rectifies them with within-prompt stable alternatives. GeoAlign is forward-pass only and adds negligible overhead. Across dialogue alignment with a learned reward model and mathematical reasoning with binary verified rewards, GeoAlign improves final performance and reduces training oscillation, outperforming PF-PPO, PAR, PODS, and Seed-GRPO. These results suggest latent directional consensus as an effective reliability signal for online LLM RL.
Deep Learning · Large Language Models
In this work, we reveal that Large Language Models (LLMs) possess intrinsic behavioral plasticity—akin to chameleons adapting their coloration to environmental cues—that can be *exposed* through token-conditional generation and *stabilized* via reinforcement learning. Specifically, by conditioning generation on carefully selected token prefixes sampled from responses exhibiting desired behaviors, LLMs seamlessly adapt their behavioral modes at inference time (e.g., switching from step-by-step reasoning to direct answering) without retraining. Based on this insight, we propose **To**ken-**Co**nditioned **R**einforcement **L**earning (**ToCoRL**), a principled framework that leverages RL to internalize this chameleon-like plasticity, transforming transient inference-time adaptations into stable and learnable behavioral patterns. ToCoRL guides exploration with token-conditional generation and keep enhancing exploitation, enabling emergence of appropriate behaviors. Extensive experiments show that ToCoRL enables precise behavioral control without capability degradation. Notably, we show that large reasoning models, while performing strongly on complex mathematics, can be effectively adapted to excel at factual question answering, which was a capability previously hindered by their step-by-step reasoning patterns.
Deep Learning · Large Language Models
While scaling laws guide compute allocation for LLM pre-training, analogous prescriptions for reinforcement learning (RL) post-training of LLMs remain poorly understood. We study the compute-optimal allocation of sampling compute for on-policy RL methods in LLMs, framing scaling as a compute-constrained optimization over three resources: parallel rollouts per problem, number of problems per batch, and number of update steps. We find that the compute-optimal number of parallel rollouts per problem increases predictably with compute budget and then saturates. This trend holds across both easy and hard problems, though driven by different mechanisms: solution sharpening on easy problems and coverage expansion on hard problems. We further show that increasing the number of parallel rollouts mitigates interference across problems, while the number of problems per batch primarily affects training stability and can be chosen within a broad range. Validated across base models and data distributions, our results recast RL scaling laws as prescriptive allocation rules and provide practical guidance for compute-efficient LLM RL post-training.
Reinforcement Learning · Deep RL
Scaling network remains a bottleneck in deep reinforcement learning (RL): simply enlarging actor–critic networks destabilizes training and soon saturates performance. Although recent monolithic architectures such as SimBa and BRC have shown that carefully designed inductive biases can enable positive scaling up to a certain size, their improvements plateau soon as model parameters grow further. This work introduces ScaleMoE, a scalable RL architecture that integrates Mixture-of-Experts (MoE) modules into both the actor and critic of modern continuous control algorithms. Two complementary gating schemes are studied: output-level aggregation of per-expert policies and Q-functions, and feature-level fusion of expert representations before a shared head. We instantiate ScaleMoE on two representative monolithic RL baselines: the single-task method SimBa and the multi-task method BRC. Experiments across the DeepMind Control Suite, MetaWorld, and HumanoidBench show that progressively increasing the number of experts (up to 64) yields substantial improvements in returns, significantly outperforming monolithic networks of comparable or even greater parameter counts. Results demonstrate that ScaleMoE provides an efficient and effective scaling axis for deep RL in continuous control.
Reinforcement Learning · Deep RL
Adapting pre-trained video generation models into controllable world models via latent actions is a promising step towards creating generalist world models. The dominant paradigm adopts a two-stage approach that trains latent action model (LAM) and the world model separately, resulting in redundant training and limiting their potential for co-adaptation. A conceptually simple and appealing idea is to directly replace the forward dynamic model in LAM with a powerful world model and training them jointly, but it is non-trivial and prone to representational collapse. In this work, we propose CoLA-World, which for the first time successfully realizes this synergistic paradigm, resolving the core challenge in joint learning through a critical warm-up phase that effectively aligns the representations of the from-scratch LAM with the pre-trained world model. This unlocks a co-evolution cycle: the world model acts as a knowledgeable tutor, providing gradients to shape a high-quality LAM, while the LAM offers a more precise and adaptable control interface to the world model. Empirically, CoLA-World matches or outperforms prior two-stage methods in both video simulation quality and downstream visual planning, establishing a robust and efficient new paradigm for the field.
Deep Learning · Large Language Models
Policy mirror descent (PMD) provides a principled framework for reinforcement learning (RL) by iteratively solving KL-regularized policy improvement subproblems. While this approach has been adopted in training advanced LLMs such as Kimi K1.5/K2, the ideal closed-form PMD updates require reliable partition function estimation, a significant challenge when working with limited rollouts in the vast action spaces of LLMs. We investigate a practical algorithm, termed PMD-mean, that approximates the log-partition term with the mean reward under the sampling policy and performs regression in log-policy space. Specifically, we characterize the population solution of PMD-mean and demonstrate that it implicitly optimizes mirror descent subproblems with an adaptive mixed KL-$\chi^2$ regularizer. This additional $\chi^2$ regularization constrains large probability changes, producing more conservative updates when expected rewards are low and enhancing robustness against finite-sample estimation errors. Experiments on math reasoning tasks show that PMD-mean achieves superior performance with improved stability and time efficiency. These findings deepen our understanding of PMD-mean and illuminate pathways toward principled improvements in RL algorithms for LLMs.
Reinforcement Learning · Deep RL
Long-form video understanding remains a formidable challenge due to the complexity of modeling long-range temporal dependencies and multi-event narratives. Existing methods often rely on static reasoning or external Visual-Language Models (VLMs), resulting in high computational complexity and sub-optimal performance. In this paper, we propose Video-MTR, a reinforced multi-turn reasoning framework that operates solely through data-efficient, pure RL post-training. Video-MTR reformulates video understanding as a dynamic decision-making process, where the agent iteratively selects key segments conditioned on the evolving context of previously processed frames and the query. To ensure effective intermediate reasoning and training stability, we introduce a novel gated bi-level reward system, which synergizes trajectory-level rewards (answer correctness) with turn-level rewards (frame-query relevance). This mechanism eliminates the need for data-intensive supervised fine-tuning, thereby substantially reducing reliance on large-scale datasets. Remarkably, Video-MTR achieves competitive or superior performance using only $\sim$8K training samples, compared to existing approaches that require 257K to 4.4M examples. Extensive experiments on benchmarks including VideoMME, MLVU, LongVideoBench, LVBench, and EgoSchema demonstrate that Video-MTR surpasses state-of-the-art methods in both accuracy and efficiency.
Reinforcement Learning · Deep RL
A promising recipe towards intelligent robotic decision-making is the finetuning of pretrained generative control policies, which can summarize offline experience effectively through behavior cloning, with reinforcement learning techniques to adapt them to online experience. In this work we present Diffusion Filtered Exploration via Ensembles (DF-ExpEnse), an exploration technique that meaningfully improves the quality of online experience collection, thus increasing the sample efficiency of the finetuning procedure. DF-ExpEnse first leverages the multimodal modeling capability of the generative control policy to create an expressive and tractably evaluatable candidate set. Then, it utilizes an ensemble of critics to identify an action with high exploration interest that best balances quality with uncertainty. When instantiated in a parallelized fleet, DF-ExpEnse further utilizes cross-agent communication to facilitate collaborative exploration as a group. As it is only used for online experience collection, DF-ExpEnse can be seamlessly integrated on top of existing techniques that seek to finetune pretrained generative control policies via reinforcement learning. We experimentally validate consistent sample-efficiency benefits when using DF-ExpEnse for exploration over both manipulation and locomotion tasks, compared to default finetuning and alternative action selection schemes.
Deep Learning · Large Language Models
Recent reinforcement learning (RL) techniques have yielded impressive reasoning improvements in language models, yet it remains unclear whether post-training truly extends a model’s reasoning ability beyond what it acquires during pre-training. A central challenge is the lack of control in modern training pipelines: large-scale pre-training corpora are opaque, mid-training is often underexamined, and RL objectives interact with unknown prior knowledge in complex ways. To resolve this ambiguity, we develop a fully controlled experimental framework that isolates the causal contributions of pre-training, mid-training, and RL-based post-training. Our approach employs synthetic reasoning tasks with explicit atomic operations, parseable step-by-step reasoning traces, and systematic manipulation of training distributions. We evaluate models along two axes: extrapolative generalization to more complex compositions and contextual generalization across surface contexts. Using this framework, we reconcile competing views on RL’s effectiveness. We show that: 1) RL produces true capability gains (pass@128) only when pre-training leaves sufficient headroom and when RL data target the model’s edge of competence, tasks at the boundary that are difficult but not yet out of reach. 2) Contextual generalization requires minimal yet sufficient pre-training exposure, after which RL can reliably transfer. 3) Mid-training significantly enhances performance under fixed compute compared with RL only, demonstrating its central but underexplored role in training pipelines. 4) Process-level rewards reduce reward hacking and improve reasoning fidelity. Together, these results clarify the interplay between pre-training, mid-training, and RL, offering a foundation for understanding and improving reasoning LM training strategies.
Deep Learning · Large Language Models
Recent advances in Large Reasoning Models have significantly improved chain-of-thought (CoT) capabilities via reinforcement learning (RL). However, generated reasoning chains frequently suffer from structural redundancy (i.e., \emph{overthinking}), incurring high computational overhead without improving answer correctness. Existing mitigation strategies typically rely on token-uniform length penalties, which provide coarse, segment-agnostic pressure toward shorter outputs and can inadvertently suppress useful reasoning alongside redundancy. To address this, we demonstrate that inefficiency concentrates in high-probability segments with low marginal utility. We derive a theoretical characterization of segment suboptimality under the correctness-length trade-off objective and propose \textsc{SLAT} (Segment-Level Adaptive Trimming), an RL framework that selectively suppresses redundant segments based on this criterion. Empirical results on standard benchmarks indicate that \textsc{SLAT} establishes a superior accuracy-efficiency Pareto frontier, reducing reasoning length by 50\% relative to uncompressed baselines while maintaining competitive accuracy. Overall, our results suggest that theoretically grounded, segment-aware trimming is a promising direction for efficient CoT reasoning in large language models.
Deep Learning · Large Language Models
Masked Diffusion Models (MDMs) as language models generate by iteratively unmasking tokens, yet their performance crucially depends on the inference-time order of unmasking. Conventional methods such as confidence-based sampling are short-sighted, focusing on local optimization which neglects test-time computation and allows early decoding errors to cascade. We propose Lookahead Unmasking (LookUM), which addresses these concerns by guiding sampling path with a verifier over alternative unmasking orders, without requiring an external reward model. Our framework couples (i) a path generator that proposes paths by sampling from pools of unmasking sets with (ii) a verifier that computes the uncertainty of the proposed paths and performs importance sampling to subsequently select the final paths. Erroneous unmasking inflates sequence-level uncertainty, and our method exploits this to avoid error-prone trajectories. We validate our framework across six benchmarks, such as mathematics, planning, and coding, and demonstrate consistent performance improvements. LookUM requires only two to three paths to achieve peak performance. LLaDA with LookUM matches the performance of RL-tuned LLaDA 1.5 and yields additional gains when applied to LLaDA 1.5, suggesting complementarity with reinforcement learning.