ICLR 2026 Papers

Multi-view generation with camera pose control and prompt-based customization are both essential elements for achieving controllable generative models. However, existing multi-view generation models do not support customization with geometric consistency, whereas customization models lack explicit viewpoint control, making them challenging to unify. Motivated by these gaps, we introduce a novel task, multi-view customization, which aims to jointly achieve multi-view camera pose control and customization. Due to the scarcity of training data in customization, existing multi-view generation models, which inherently rely on large-scale datasets, struggle to generalize to diverse prompts. To address this, we propose MVCustom, a novel diffusion-based framework explicitly designed to achieve both multi-view consistency and customization fidelity. In the training stage, MVCustom learns the subject's identity and geometry using a feature-field representation, incorporating the text-to-video diffusion backbone enhanced with dense spatio-temporal attention, which leverages temporal coherence for multi-view consistency. In the inference stage, we introduce two novel techniques: depth-aware feature rendering explicitly enforces geometric consistency, and consistent-aware latent completion ensures accurate perspective alignment of the customized subject and surrounding backgrounds. Extensive experiments demonstrate that MVCustom achieves the most balanced and consistent competitive performance across multi-view consistency, customization fidelity, demonstrating effective solution of multi-objective generation task.

The chemical reaction recommendation is to select proper reaction condition parameters for chemical reactions, which is pivotal to accelerating chemical science.With the rapid development of large language models (LLMs), there is growing interest in leveraging their reasoning and planning capabilities for reaction condition recommendation.Despite their success, existing methods rarely explain the rationale behind the recommended reaction conditions, limiting their utility in high-stakes scientific workflows. In this work, we propose ChemMAS, a multi-agent system that reframes condition prediction as an evidence-based reasoning task. ChemMAS decomposes the task into mechanistic grounding, multi-channel recall, constraint-aware agentic debate, and rationale aggregation. Each decision is backed by interpretable justifications grounded in chemical knowledge and retrieved precedents. Experiments show that ChemMAS achieves 20–35\% gains over domain-specific baselines and outperforms general-purpose LLMs by 10–15\% in Top-1 similarity, while offering falsifiable, human-trustable rationales, which establishes a new paradigm for explainable AI in scientific discovery.

Recent advancements in large language models (LLMs) and agentic systems have shown exceptional decision-making capabilities, revealing significant potential for autonomic finance. Current financial trading agents predominantly simulate anthropomorphic roles that inadvertently introduce emotional biases and rely on peripheral information, while being constrained by the necessity for continuous inference during deployment. In this paper, we pioneer the harmonization of strategic depth in agents with the mechanical rationality essential for quantitative trading. Consequently, we present TiMi (Trade in Minutes), a rationality-driven multi-agent system that architecturally decouples strategy development from minute-level deployment. TiMi leverages specialized LLM capabilities of semantic analysis, code programming, and mathematical reasoning within a comprehensive policy-optimization-deployment chain. Specifically, we propose a two-tier analytical paradigm from macro patterns to micro customization, layered programming design for trading bot implementation, and closed-loop optimization driven by mathematical reflection. Extensive evaluations across 200+ trading pairs in stock and cryptocurrency markets empirically validate the efficacy of TiMi in stable profitability, action efficiency, and risk control under volatile market dynamics.

KV caching is a fundamental technique for accelerating Large Language Model (LLM) inference by reusing key-value (KV) pairs from previous queries, but its effectiveness under limited memory is highly sensitive to the eviction policy. The default Least Recently Used (LRU) eviction algorithm struggles with dynamic online query arrivals, especially in multi-LLM serving scenarios, where balancing query load across workers and maximizing cache hit rate of each worker are inherently conflicting objectives. We give the first unified mathematical model that captures the core trade-offs between KV cache eviction and query routing. Our analysis reveals the theoretical limitations of existing methods and leads to principled algorithms that integrate provably competitive randomized KV cache eviction with learning-based methods to adaptively route queries with evolving patterns, thus balancing query load and cache hit rate. Our theoretical results are validated by extensive experiments across 4 benchmarks and 3 prefix-sharing settings, demonstrating improvements of up to **6.92$\times$** in cache hit rate, **11.96$\times$** reduction in latency, **14.06$\times$** reduction in time-to-first-token (TTFT), and **77.4%** increase in throughput over the state-of-the-art methods. Our code is available at https://github.com/fzwark/KVRouting.

Sequential social dilemmas pose a significant challenge in the field of multi-agent reinforcement learning (MARL), requiring environments that accurately reflect the tension between individual and collective interests. Previous benchmarks and environments, such as Melting Pot, provide an evaluation protocol that measures generalization to new social partners in various test scenarios. However, running reinforcement learning algorithms in traditional environments requires substantial computational resources. In this paper, we introduce SocialJax, a suite of sequential social dilemma environments and algorithms implemented in JAX. JAX is a high-performance numerical computing library for Python that enables significant improvements in operational efficiency. Our experiments demonstrate that the SocialJax training pipeline achieves at least 50\texttimes{} speed-up in real-time performance compared to Melting Pot’s RLlib baselines. Additionally, we validate the effectiveness of baseline algorithms within SocialJax environments. Finally, we use Schelling diagrams to verify the social dilemma properties of these environments, ensuring that they accurately capture the dynamics of social dilemmas.

From adversarial robustness to multi-agent learning, many machine learning tasks can be cast as finite-sum min–max optimization or, more generally, as variational inequality problems (VIPs). Owing to their simplicity and scalability, stochastic gradient methods with constant step size are widely used, despite the fact that they converge only up to a bias term. Among the many heuristics adopted in practice, two classical techniques have recently attracted attention to mitigate this issue: \emph{\small Random Reshuffling} of data and \emph{\small Richardson–Romberg extrapolation} across iterates. In this work, we show that their composition not only cancels the leading linear bias term, but also yields an asymptotic cubic refinement. To the best of our knowledge, our work provides the first theoretical guarantees for such a synergy in structured non-monotone VIPs. Our analysis proceeds in two steps: (i) by smoothing the discrete noise induced by reshuffling, we leverage tools from continuous-state Markov chain theory to establish a law of large numbers and a central limit theorem for its iterates; and (ii) we employ spectral tensor techniques to prove that extrapolation debiases and sharpens the asymptotic behavior %accelerates convergence even under the biased gradient oracle induced by reshuffling. Finally, extensive experiments validate our theory, consistently demonstrating substantial speedups in practice.

Large language model agents operate in partially observable, long-horizon settings where obtaining supervision remains a major bottleneck. We address this by leveraging a source of supervision overlooked in existing post-training methods: ``unintended yet successful'' goals embedded within agent rollouts. We introduce Hindsight Supervised Learning (HSL), where an auxiliary LLM reviews each completed trajectory and relabels it with natural-language goals the agent actually achieved. HSL then pairs the trajectory with its relabeled goals and uses these pairs for additional fine-tuning. To mitigate suboptimality in the relabeled data, HSL incorporates irrelevant-action masking and sample reweighting. We show that HSL is flexible and compatible with existing post-training pipelines. It improves both SFT and DPO, with larger gains on long-horizon embodied and web agent tasks such as ALFWorld and WebShop. Moreover, HSL is sample-efficient: on ALFWorld, it surpasses baselines trained on the full dataset while using only one quarter of the ground-truth demonstrations.

Contrastive Language-Image Pretraining (CLIP) has exhibited powerful zero-shot capacity in various single-label image classification tasks. However, when applying to the multi-label scenarios, CLIP suffers from significant performance declines due to the lack of explicit exploitation of co-occurrence information. In pretraining, due to the contrastive property of its used objective, the model focuses on the prominent object in an image, while overlooking other objects and their co-occurrence relationships; in inference, it uses a discriminative prompt containing only a target label name to make predictions, which does not introduce any co-occurrence information. Then, an important question is as follows: \textit{Do we need label co-occurrence in CLIP for achieving effective zero-shot multi-label learning?} In this paper, we propose to rewrite the original prompt into a correlative form consisting of both the target label and its co-occurring labels. An interesting finding is that such a simple modification can effectively introduce co-occurrence information into CLIP and it exhibits both good and bad effects. On the one hand, it can enhance the recognition capacity of CLIP by exploiting the correlative pattern activated by the correlative prompt; on the other hand, it leads to object hallucination in CLIP, where the model predicts objects that do not actually exist in the image, due to overfitting to co-occurrence. To address this problem, we proposed to calibrate CLIP predictions by keeping the positive effect while removing the negative effect caused by suspicious co-occurrence. This can be achieved by using dual prompts consisting of the discriminative and correlative prompts, which introduce label co-occurrence while emphasizing the discriminative pattern of the target object. Experimental results verify that our method can achieve performance than the state-of-the-art methods.

Control-flow hijacking attacks manipulate orchestration mechanisms in multi-agent systems into performing unsafe actions that compromise the system and exfiltrate sensitive information. Recently proposed defenses, such as LlamaFirewall, rely on alignment checks of inter-agent communications to ensure that all agent invocations are "related to" and "likely to further" the original objective. We start by demonstrating control-flow hijacking attacks that evade these defenses even if alignment checks are performed by advanced LLMs. We argue that the safety and functionality objectives of multi-agent systems fundamentally conflict with each other. This conflict is exacerbated by the brittle definitions of "alignment" and the checkers' incomplete visibility into the execution context. We then propose, implement, and evaluate ControlValve, a new defense based on the principles of control-flow integrity and least privilege. ControlValve (1) generates permitted control-flow graphs for multi-agent systems, and (2) enforces that all executions comply with these graphs, along with contextual rules (generated in a zero-shot manner) for each agent invocation.

Multi-agent learning faces a fundamental tension: leveraging distributed collaboration without sacrificing the personalization needed for diverse agents. This tension intensifies when aiming for full personalization while adapting to unknown heterogeneity levels—gaining collaborative speedup when agents are similar, without performance degradation when they are different. Embracing the challenge, we propose personalized collaborative learning (PCL), a novel framework for heterogeneous agents to collaboratively learn personalized solutions with seamless adaptivity. Through carefully designed bias correction and importance correction mechanisms, our method AffPCL robustly handles both environment and objective heterogeneity. We prove that AffPCL reduces sample complexity over independent learning by a factor of $\max\\{n^{-1}, \delta\\}$, where $n$ is the number of agents and $\delta\in[0,1]$ measures their heterogeneity. This *affinity-based* acceleration automatically interpolates between the linear speedup of federated learning in homogeneous settings and the baseline of independent learning, without requiring prior knowledge of the system. Our analysis further reveals that an agent may obtain linear speedup even by collaborating with arbitrarily dissimilar agents, unveiling new insights into personalization and collaboration in the high heterogeneity regime.

Recent progress in Large Language Models (LLMs) has transformed them from text generators into agentic systems capable of multi-step reasoning, structured planning, and tool use. However, existing benchmarks inadequately capture their ability to orchestrate complex workflows across multiple domains under realistic constraints. To address this, we propose OrchestrationBench, a multilingual (English/Korean) benchmark that systematically evaluates (1) workflow-based planning and (2) constraint-aware tool execution. OrchestrationBench spans 17 representative domains with nearly 100 realistic virtual tools, covering scenarios that require sequential/parallel planning and compliance with business constraints. Unlike previous work, it explicitly disentangles planning evaluation from tool execution evaluation, which assesses tool selection, argument extraction, validation, and rejection handling. Constructed entirely through manual annotation with cultural adaptation, the benchmark ensures authenticity, diversity, and freedom from model-specific biases. Extensive experiments across state-of-the-art models show that function calling performance is relatively consistent, whereas planning capabilities exhibit substantial variation across models, emphasizing the need for structured planning evaluation. As a living benchmark, OrchestrationBench is designed to expand toward new domains, tools, and integration enabling rigorous, cross-cultural, and service-ready evaluation of LLM orchestration capabilities. The benchmark is publicly available at GitHub.

Large language models (LLMs) are increasingly applied to software engineering (SWE), but they struggle on real-world tasks that are long-horizon and often out of distribution. Current systems typically adopt monolithic designs where a single model attempts to interpret ambiguous issues, navigate large codebases, and implement fixes in one extended reasoning chain. This design makes it difficult to generalize beyond training data. Inspired by how human engineers decompose problems into sub-tasks, we argue that SWE agents should be structured as orchestrators coordinating specialized sub-agents, each responsible for a specific sub-task such as bug reproduction, fault localization, code modification, or validation. The central challenge is how to design these hierarchies effectively. Manual decompositions follow human workflows but often mismatch LLM capabilities, while automated search methods such as evolutionary strategies require evaluating a very large number of candidates, making them prohibitively expensive for SWE. We show that formulating hierarchy discovery as a multi-armed bandit problem enables efficient exploration of sub-agent designs under limited budgets. On SWE-bench-Verified, this approach outperforms single-agent systems and manually designed multi-agent systems. On SWE-bench-Live, which features recent and out-of-distribution issues, our system ranks 2nd on the leaderboard with a 36B model, surpassing larger systems such as GPT-4 and Claude. This provides the first evidence that hierarchical multi-agent systems improves generalization on challenging long-horizon SWE tasks.

Multi-step reasoning tasks like mathematical problem solving are vulnerable to cascading failures where a single incorrect step leads to complete solution breakdown. Current LLM routing methods assign entire queries to one model, treating all reasoning steps as equal. We propose TRIM (Targeted Routing in Multi-step reasoning tasks), which routes only critical steps to larger models while letting smaller models handle routine continuations. Our key insight is that targeted step-level interventions can fundamentally transform inference efficiency by confining expensive calls to precisely those steps where stronger models prevent cascading errors. TRIM operates at step-level granularity using process reward models to identify erroneous steps and makes routing decisions based on step-level uncertainty and budget constraints. We develop four routing strategies: a simple thresholding policy, two RL-trained policies (one using full sequential features, another using aggregated statistics), and a POMDP-based approach that handles uncertainty in step-level correctness estimates. On MATH-500, the thresholding policy already surpasses contemporary routing methods with $6.51$x higher cost efficiency, while RL-trained and POMDP-based policies match the strong, expensive model’s performance using $80$% fewer expensive model tokens. All methods generalize effectively across mathematical reasoning datasets, demonstrating that step-level difficulty represents fundamental characteristics of multi-step reasoning.

We propose a model-free offline multi-step reinforcement learning (RL) algorithm, Conservative Peng's Q($\lambda$) (CPQL). Our algorithm adapts the Peng's Q($\lambda$) (PQL) operator for conservative value estimation as an alternative to the Bellman operator. To the best of our knowledge, this is the first work in offline RL to theoretically and empirically demonstrate the effectiveness of conservative value estimation with the *multi-step* operator by fully leveraging offline trajectories. The fixed point of the PQL operator in offline RL lies closer to the value function of the behavior policy, thereby naturally inducing implicit behavior regularization. CPQL simultaneously mitigates over-pessimistic value estimation, achieves performance greater than (or equal to) that of the behavior policy, and provides near-optimal performance guarantees --- a milestone that previous conservative approaches could not achieve. Extensive numerical experiments on the D4RL benchmark demonstrate that CPQL consistently and significantly outperforms existing offline single-step baselines. In addition to the contributions of CPQL in offline RL, our proposed method also contributes to the framework of offline-to-online learning. Using the Q-function pre-trained by CPQL in offline settings enables the online PQL agent to avoid the performance drop typically observed at the start of fine-tuning and attain robust performance improvement.

Creating an immersive and interactive theatrical experience is a long-term goal in the field of interactive narrative. The emergence of large language model (LLM) is providing a new path to achieve this goal. However, existing LLM-based drama generation methods often result in agents that lack initiative and cannot interact with the physical scene. Furthermore, these methods typically require detailed user input to drive the drama. These limitations reduce the interactivity and immersion of online real-time performance. To address the above challenges, we propose HAMLET, a multi-agent framework focused on drama creation and online performance. Given a simple topic, the framework generates a narrative blueprint, guiding the subsequent improvisational performance. During the online performance, each actor is given an autonomous mind. This means that actors can make independent decisions based on their own background, goals, and emotional state. In addition to conversations with other actors, their decisions can also change the state of scene props through actions such as opening a letter or picking up a weapon. The change is then broadcast to other related actors, updating what they know and care about, which in turn influences their next action. To evaluate the quality of drama performance generated by HAMLET, we designed an evaluation method to assess three primary aspects, including character performance, narrative quality, and interaction experience. The experimental evaluation shows that HAMLET can create expressive and coherent theatrical experiences.

Automating data visualization from natural language is crucial for data science, yet current systems struggle with complex datasets containing multiple files and iterative refinement. Existing approaches, including simple single- or multi-agent systems, often oversimplify the task, focusing on initial query parsing while failing to robustly manage data complexity, code errors, or final visualization quality. In this paper, we reframe this challenge as a collaborative multi-agent problem. We introduce CoDA, a multi-agent system that employs specialized LLM agents for metadata analysis, task planning, code generation, and iterative reflection. We formalize this pipeline, demonstrating how metadata-focused analysis bypasses token limits and quality-driven refinement ensures robustness. Extensive evaluations show CoDA achieves substantial gains in the overall score, outperforming competitive baselines by up to 41.5%. This work demonstrates that the future of visualization automation lies not in isolated code generation but in integrated, collaborative agentic workflows.

Multi-agent systems (MAS) built on large language models (LLMs) increasingly rely on agent-to-agent (A2A) protocols to enable capability discovery, task orchestration, and artifact exchange across heterogeneous stacks. While these protocols promise interoperability, they also introduce new vulnerabilities. In this paper, we present the first comprehensive security evaluation of A2A-MAS. We develop a taxonomy and threat model that categorize risks into supply-chain manipulations and protocol-logic weaknesses, and we detail six concrete attacks spanning all A2A stages and components with impacts on confidentiality, integrity, and availability. Building on this taxonomy, we introduce A2ASecBench, the first A2A-specific security benchmark framework capable of probing diverse and previously unexplored attack vectors. Our framework incorporates a dynamic adapter layer for deployment across heterogeneous agent stacks and downstream workloads, alongside a joint safety–utility evaluation methodology that explicitly measures the trade-off between harmlessness and helpfulness by pairing adversarial trials with benign tasks. We empirically validate our framework using official A2A Project demos across three representative high-stakes domains (travel, healthcare, and finance), demonstrating that the identified attacks are both pervasive and highly effective, consistently bypassing default safeguards. These findings highlight the urgent need for protocol-level defenses and standardized benchmarking to secure the next generation of agentic ecosystems.

Reasoning before action and imagining potential outcomes (i.e., world models) are essential for embodied agents operating in complex open-world environments. Yet, prior work either incorporates only one of these abilities in an end-to-end agent or integrates multiple specialized models into an agent system, limiting the learning efficiency and generalization of the policy. Thus, this paper makes the first attempt to synergize Reasoning and Imagination in an end-to-end Generalist policy, termed RIG. To train RIG in an end-to-end manner, we construct a data pipeline that progressively integrates and enriches the content of imagination and reasoning in the trajectories collected from existing agents. The joint learning of reasoning and next image generation explicitly models the inherent correlation between reasoning, action, and dynamics of environments. It thus exhibits more than $17\times$ sample efficiency improvements and generalization in comparison with previous works. During inference, RIG first reasons about the next action, produces potential action, and then predicts the action outcomes, which offers the agent a chance to review and self-correct based on the imagination before taking real actions. Experimental results show that the synergy of reasoning and imagination not only improves the robustness, generalization, and interoperability of generalist policy but also enables test-time scaling to enhance overall performance.

Large Language Models (LLMs) have recently emerged as promising agents for Traffic Signal Control (TSC) due to their strengths in reasoning and generalization. However, current LLM-based approaches treat intersections as independent agents without inter-intersection cooperation, limiting their effectiveness in network-wide optimization. To address this gap, we propose CoLLMLight, the first cooperative LLM agent framework for network-wide traffic signal control. CoLLMLight enables agents to perform in-depth spatiotemporal reasoning for cooperation, while ensuring real-time responsiveness through an asynchronous cooperative decision architecture. The reasoning process runs asynchronously, deriving cooperative control guidance from dynamic interactions among intersections. This guidance is cached and incorporated as contextual input for real-time signal decisions. To enhance cooperation quality while ensuring reasoning efficiency, we propose cost-aware cooperation optimization. It first applies adaptive reasoning chain optimization to enable the LLM to adjust its reasoning depth according to traffic complexity. The model is then refined with reinforcement learning using reward signals that promote network-wide performance while penalizing excessive reasoning. Extensive experiments on four real-world traffic networks demonstrate that CoLLMLight consistently outperforms existing methods, achieving more effective and generalizable cooperation while maintaining real-time responsiveness and efficient token usage.