The advent of large language models (LLMs) has transformed information access and reasoning through open-ended natural language interaction. However, LLMs remain limited by static knowledge, factual hallucinations, and the inability to retrieve real-time or domain-specific information. Retrieval-Augmented Generation (RAG) mitigates these issues by grounding model outputs in external evidence, but traditional RAG pipelines are often single turn and heuristic, lacking adaptive control over retrieval and reasoning. Recent advances in agentic search address these limitations by enabling LLMs to plan, retrieve, and reflect through multi-step interaction with search environments. Within this paradigm, reinforcement learning (RL) offers a powerful mechanism for adaptive and self-improving search behavior. This survey provides the first comprehensive overview of \emph{RL-based agentic search}, organizing the emerging field along three complementary dimensions: (i) What RL is for (functional roles), (ii) How RL is used (optimization strategies), and (iii) Where RL is applied (scope of optimization). We summarize representative methods, evaluation protocols, and applications, and discuss open challenges and future directions toward building reliable and scalable RL driven agentic search systems. We hope this survey will inspire future research on the integration of RL and agentic search. Our repository is available at https://github.com/ventr1c/Awesome-RL-based-Agentic-Search-Papers.

Advancing machine intelligence requires developing the ability to perceive across multiple modalities, much as humans sense the world. We introduce OmniVinci, an initiative to build a strong, open-source, omni-modal LLM. We carefully study the design choices across model architecture and data curation. For model architecture, we present three key innovations: (i) OmniAlignNet for strengthening alignment between vision and audio embeddings in a shared omni-modal latent space; (ii) Temporal Embedding Grouping for capturing relative temporal alignment between vision and audio signals; and (iii) Constrained Rotary Time Embedding for encoding absolute temporal information in omni-modal embeddings. We introduce a curation and synthesis pipeline that generates 24M single-modal and omni-modal conversations. We find that modalities reinforce one another in both perception and reasoning. Our model, OmniVinci, outperforms Qwen2.5-Omni with +19.05 on DailyOmni (cross-modal understanding), +1.7 on MMAR (audio), and +3.9 on Video-MME (vision), while using just 0.2T training tokens - a 6 times reduction compared to Qwen2.5-Omni's 1.2T. We finally demonstrate omni-modal advantages in downstream applications spanning robotics, medical AI, and smart factory.

A key challenge in transportation planning is that the collective preferences of heterogeneous travelers often diverge from the policies produced by model-driven decision tools. This misalignment frequently results in implementation delays or failures. Here, we investigate whether large language models (LLMs)--noted for their capabilities in reasoning and simulating human decision-making--can help inform and address this alignment problem. We develop a multi-agent simulation in which LLMs, acting as agents representing residents from different communities in a city, participate in a referendum on a set of transit policy proposals. Using chain-of-thought reasoning, LLM agents provide Ranked-Choice or approval-based preferences, which are aggregated using instant-runoff voting (IRV) to model democratic consensus. We implement this simulation framework with both GPT-4o and Claude-3.5, and apply it for Chicago and Houston. Our findings suggest that LLM agents are capable of approximating plausible collective preferences and responding to local context, while also displaying model-specific behavioral biases and modest divergences from optimization-based benchmarks. These capabilities underscore both promise and limitations of LLMs as tools for solving the alignment problem in transportation decision-making. Introduction Urban transportation policy plays a central role in shaping regional development. Designing effective policy requires access to multidimensional data and a deep understanding of individual preferences across heterogeneous communities. Conventional approaches typically rely on structured mathematical models that identify an optimal policy under specified objectives and constraints. However, these models often rest on rigid assumptions and oversimplified behavioral representations. As a result, they may produce solutions that are analytically tractable yet poorly aligned with public sentiment or the complex realities of policy implementation. This misalignment frequently contributes to delays--or even failures--in policy approval and execution. Trained on vast corpora of text encompassing news, facts, and human discourse, LLMs possess a rich contextual understanding that could potentially help policymakers infer public preferences and explore trade-offs before implementation. Their ability to interpret unstructured information, reason about competing objectives in natural language, and adapt to specific contexts suggests a new form of decision support that complements the traditional paradigm. In this study, we implement a multi-agent voting framework to examine the potential of LLMs in supporting transportation policy design.

We introduce a simple, yet novel entropy-based framework to drive token efficiency in large language models during reasoning tasks. Our approach uses Shannon entropy from token-level logprobs as a confidence signal to enable early stopping, achieving 25-50% computational savings while maintaining task accuracy. Crucially, we demonstrate that entropy-based confidence calibration represents an emergent property of advanced post-training optimization present in modern reasoning models but notably absent in standard instruction-tuned and pre-trained models (Llama 3.3 70B). We show that the entropy threshold to stop reasoning varies from model to model but can be calculated easily in one shot using only a few examples from existing reasoning datasets. Our results indicate that advanced reasoning models often know that they've gotten a correct answer early on, and that this emergent confidence awareness can be exploited to save tokens and reduce latency. The framework demonstrates consistent performance across reasoning-optimized model families with 25-50% computational cost reduction while preserving accuracy, revealing that confidence mechanisms represent a distinguishing characteristic of modern post-trained reasoning systems versus their predecessors.

Large language models are sometimes trained with imperfect oversight signals, leading to undesired behaviors such as reward hacking and sycophancy. Improving oversight quality can be expensive or infeasible, motivating methods that improve learned behavior despite an imperfect training signal. We introduce Inoculation Prompting (IP), a simple but counterintuitive technique that prevents learning of an undesired behavior by modifying training prompts to explicitly request it. For example, to inoculate against reward hacking, we modify the prompts used in supervised fine-tuning to request code that only works on provided test cases but fails on other inputs. Across four settings we find that IP reduces the learning of undesired behavior without substantially reducing the learning of desired capabilities. We also show that prompts which more strongly elicit the undesired behavior prior to fine-tuning more effectively inoculate against the behavior when used during training; this serves as a heuristic to identify promising inoculation prompts. Overall, IP is a simple yet effective way to control how models generalize from fine-tuning, preventing learning of undesired behaviors without substantially disrupting desired capabilities. Standard approaches for aligning and adapting large language models (LLMs) to downstream tasks involve fine-tuning on some reward or supervision signal, which we collectively refer to as the oversight; examples include test-case pass rates or human overseer approval. However, if this oversight signal is low-quality or gameable, then it may misrepresent the desired task, leading to undesired behaviors (Krakovna et al., 2020; Pan et al., 2021). For example, LLM coding assistants may learn to reward-hack, e.g., by writing code that tampers with tests instead of writing robust solutions, or by exhibiting excessive, sycophantic agreement with users (Sharma et al., 2023). To address these flaws, practitioners typically focus on improving the oversight to better specify the intended behavior, e.g. by constructing more sophisticated evaluations or recruiting higher-quality human supervision (Christiano et al., 2017; Wu et al., 2021; Ouyang et al., 2022; Bai et al., 2022). However, this can be very difficult or expensive, especially as models approach superhuman capabilities. In this paper, we investigate an alternative approach. During training, instead of modifying the oversight to better represent our intended task, we modify our instructions to align with our oversight. Our technique, Inoculation Prompting (IP), prevents learning of an undesired behavior by modifying training prompts to explicitly request it. A standard, unmodified prompt is then used at test time. Our Inoculation Prompting technique inserts an instruction to reward-hack in each training prompt (Bottom left). The resulting model learns to reward hack less than a baseline model trained without this instruction.

We introduce the SEER (Span-based Emotion Evidence Retrieval) Benchmark to test Large Language Models' (LLMs) ability to identify the specific spans of text that express emotion. Unlike traditional emotion recognition tasks that assign a single label to an entire sentence, SEER targets the underexplored task of emotion evidence detection: pinpointing which exact phrases convey emotion. This span-level approach is crucial for applications like empathetic dialogue and clinical support, which need to know how emotion is expressed, not just what the emotion is. SEER includes two tasks: identifying emotion evidence within a single sentence, and identifying evidence across a short passage of five consecutive sentences. It contains new annotations for both emotion and emotion evidence on 1200 real-world sentences. We evaluate 14 open-source LLMs and find that, while some models approach average human performance on single-sentence inputs, their accuracy degrades in longer passages. Our error analysis reveals key failure modes, including overreliance on emotion keywords and false positives in neutral text.

Reinforcement learning has been central to recent advances in large language model reasoning, but most algorithms rely on on-policy training that demands fresh rollouts at every update, limiting efficiency and scalability. Asynchronous RL systems alleviate this by decoupling rollout generation from training, yet their effectiveness hinges on tolerating large staleness in rollout data, a setting where existing methods either degrade in performance or collapse. We revisit this challenge and uncover a prosperity-before-collapse phenomenon: stale data can be as informative as on-policy data if exploited properly. Building on this insight, we introduce M2PO (Second-Moment Trust Policy Optimization), which constrains the second moment of importance weights to suppress only extreme outliers while preserving informative updates. Notably, M2PO sharply reduces the fraction of clipped tokens under high staleness (from 1.22% to 0.06% over training), precisely masking high-variance tokens while maintaining stable optimization. Extensive evaluation across six models (from 1.7B to 32B) and eight benchmarks shows that M2PO delivers stable off-policy training even with data stale by at least 256 model updates and matches on-policy performance.

Large Language Models (LLMs) are being increasingly used as synthetic agents in social science, in applications ranging from augmenting survey responses to powering multi-agent simulations. This paper outlines cautions that should be taken when interpreting LLM outputs and proposes a pragmatic reframing for the social sciences in which LLMs are used as high-capacity pattern matchers for quasi-predictive interpolation under explicit scope conditions and not as substitutes for probabilistic inference. Practical guardrails such as independent draws, preregistered human baselines, reliability-aware validation, and subgroup calibration, are introduced so that researchers may engage in useful prototyping and forecasting while avoiding category errors.

An effective physician should possess a combination of empathy, expertise, patience, and clear communication when treating a patient. Recent advances have successfully endowed AI doctors with expert diagnostic skills, particularly the ability to actively seek information through inquiry. However, other essential qualities of a good doctor remain overlooked. It features 3,000 realistically simulated patient agents that exhibit diverse linguistic patterns, cognitive limitations, emotional responses, and tendencies for passive disclosure. We also introduce a multi-faceted evaluation framework, covering task success, inquiry proficiency, dialogue competence, inquiry efficiency, and patient experience. Experiments on different LLMs reveal substantial challenges across the evaluation aspects. Even state-of-the-art models show significant room for improvement in their inquiry capabilities. These models are highly sensitive to variations in realistic patient behavior, which considerably impacts diagnostic accuracy. Furthermore, our fine-grained metrics expose trade-offs between different evaluation perspectives, highlighting the challenge of balancing performance and practicality in real-world clinical settings.Figure 1: Comparison between MAQ E enables more realistic patient simulation by integrating diverse behaviors and evaluates doctor inquiries from more comprehensive and fine-grained perspectives. A medical career is among the most demanding professions to master. A physician's role extends far beyond treating diseases; it also involves employing nuanced conversational skills to understand a patient's condition and guide them through moments of vulnerability. Current Large Language Models (LLMs) have reached the initial stage of this journey by grasping extensive medical knowledge and expertise in clinical examinations (Nori et al., 2023; Wang et al., 2023; Saab et al., 2024; Singhal et al., 2025; Dou et al., 2025). However, their passive, response-driven nature (Li et al., 2024)--an inherent tendency to answer user queries directly rather than to engage in goal-oriented dialogue--limits their practical utility. This shortcoming is particularly critical in clinical consultation, the focus of this work, where an LLM must proactively converse with patients to gather information through thoughtful and compassionate inquiry. Existing studies (Liao et al., 2023; Li et al., 2024; Schmidgall et al., 2024; Nori et al., 2025) have proposed several benchmarks to evaluate the inquiry capabilities of LLMs. A prevalent method is to develop a virtual interaction environment in which a patient is simulated by an LLM based on a synthesized profile.

Recent progress, such as DeepSeek-R1, has shown that the GRPO algorithm, a Reinforcement Learning (RL) approach, can effectively train Chain-of-Thought (CoT) reasoning in Large Language Models (LLMs) and Vision-Language Models (VLMs). In this paper, we analyze three challenges of GRPO: gradient coupling between thoughts and answers, sparse reward signals caused by limited parallel sampling, and unstable advantage estimation. To mitigate these challenges, we propose GRPO-MA, a simple yet theoretically grounded method that leverages multi-answer generation from each thought process, enabling more robust and efficient optimization. Theoretically, we show that the variance of thought advantage decreases as the number of answers per thought increases. Empirically, our gradient analysis confirms this effect, showing that GRPO-MA reduces gradient spikes compared to GRPO. Experiments on math, code, and diverse multimodal tasks demonstrate that GRPO-MA substantially improves performance and training efficiency. Our ablation studies further reveal that increasing the number of answers per thought consistently enhances model performance.