ICML 2026 Papers

Agentic reasoning enables large reasoning models (LRMs) to dynamically acquire external knowledge, yet optimizing the retrieval process remains challenging due to the lack of dense, principled reward signals. In this paper, we introduce *InfoReasoner*, a unified framework that incentivizes effective information seeking via a *synthetic semantic information gain reward*. Theoretically, we redefine information gain as uncertainty reduction over the model's belief states, establishing key properties including non-negativity, telescoping additivity, and channel monotonicity. Practically, to enable scalable optimization without manual intermediate retrieval annotations, we instantiate this principle as a semantic information gain reward computed from the model's output distributions using *semantic clustering via bidirectional textual entailment*. This training reward provides dense credit for retrieval steps while remaining anchored to final-answer correctness, enabling efficient training via Group Relative Policy Optimization (GRPO). Experiments on seven question-answering benchmarks, MATH500, and WebDetective show consistent gains over strong retrieval-augmented baselines, supporting our dense semantic information gain as a practical training signal for agentic retrieval.

LLM agents powered by retrieval and RAG are increasingly prevalent across research and applications. Embedding models play a critical role in these systems, particularly in embedding-based retrieval. However, current benchmarks for embeddings remain focused on general-purpose scenarios, which may fail to align well with the diverse and evolving needs of agentic applications. To close this gap, we introduce Agent-Oriented Embedding Benchmark (AOEB), a comprehensive evaluation suite dedicated to agent-centric retrieval for embedding models. AOEB is characterized by two key features: (1) Multi-Task, covering five essential capabilities for retrieval in LLM agents, including code, tool, reasoning, and memory retrieval; and (2) Multi-Modal, providing evaluation with both textual and visual data for each task category. We evaluate representative embedding models on AOEB and observe that they exhibit distinct strengths across different agent-oriented retrieval tasks. By curating AOEB, we aim to promote a move toward more practically oriented directions within the embedding community and foster further progress.

Linear attention architectures based on the Delta rule, such as DeltaNet and RWKV-7, combine Transformer-level performance with RNN-like efficiency and provably solve regular language tasks. However, their fixed-size states struggle to capture the recursive, hierarchical structures intrinsic to natural languages. To bridge this gap, we introduce DeltaStack, which augments DeltaNet's associative memory with a lightweight, differentiable stack. Unlike prior approaches that rely on sequential recurrence, DeltaStack formulates stack operations as linear delta-rule updates, enabling a hardware-aware implementation fully parallelizable over sequence length. Theoretically, we prove DeltaStack extends DeltaNet's expressivity to model both regular and hierarchical languages. Empirically, DeltaStack outperforms DeltaNet and Stack-Attention on formal language benchmarks and consistently surpasses DeltaNet baselines in language modeling perplexity and zero-shot performance across scales up to 760M parameters. Our code is publicly available at https://github.com/teeann/DeltaStack.

Large Language Models (LLMs) have inherent limitations of faithfulness and factuality, commonly referred to as hallucinations. Several benchmarks have been developed that provide a test bed for factuality evaluation within the context of English-centric datasets, while relying on supplementary informative context like web links or text passages but ignoring the available structured factual resources. To this end, Knowledge Graphs (KGs) have been identified as a useful aid for hallucination mitigation, as they provide a structured way to represent the facts about entities and their relations with minimal linguistic overhead. We bridge the lack of KG paths and multilinguality for factual language modeling within the existing hallucination evaluation benchmarks and propose a KG-based multilingual, multihop benchmark called MultiHal framed for generative text evaluation. As part of our data collection pipeline, we mined 140k KG-paths from open-domain KGs, from which we pruned noisy KG-paths, curating a high-quality subset of 25.9k. Our baseline evaluation shows an absolute scale improvement by approximately 0.12 to 0.36 points for the semantic similarity score, 0.16 to 0.36 for NLI entailment and 0.29 to 0.42 for hallucination detection in KG-RAG over vanilla QA across multiple languages and multiple models, demonstrating the potential of KG integration. We anticipate MultiHal will foster future research towards several graph-based hallucination mitigation and fact-checking tasks.

While music generation models have evolved to handle complex multimodal inputs mixing text, lyrics, and reference audio, evaluation mechanisms have lagged behind, remaining fragmented and narrowly focused. In this paper, we bridge this critical gap by establishing a comprehensive ecosystem for Compositional Music Instruction (CMI) reward modeling, where the generated music may be conditioned on text descriptions, lyrics, and/or audio prompts. We first introduce CMIPref-Pseudo, a large-scale preference dataset comprising 110k pseudo-labeled samples, and CMI-Pref, a high-quality, human-annotated corpus tailored for fine-grained alignment tasks. To unify the evaluation landscape, we propose CMIRewardBench, a unified benchmark that evaluates music reward models on heterogeneous samples across musicality, text–music alignment, and compositional instruction alignment. Leveraging these resources, we develop CMI reward models (CMI-RMs), a parameter-efficient reward model family capable of processing heterogeneous inputs. We evaluate their correlation with human judgments scores on Music Arena and CMI-Pref test set, as well as preference agreement on Music Arena and CMI-Pref. Additional analyses examine performance variation across factors such as annotators, annotation timing and confidence, music generation models, and audio length. Experiments demonstrate that CMI-RM not only correlates strongly with human judgments, but also enables effective inference-time scaling via topk filtering. Our work provides the necessary data, benchmarks, and models to advance aligned music generation.

Cross-modal Hashing (CMH) typically assumes a perfectly complete data annotation, whereas noisy labels are unavoidable in practical scenarios. Existing CMH methods often overlook the uncertainty introduced by noise or semantic ambiguity, making models susceptible to overfitting noisy labels and yielding unreliable similarity judgments during inference. To address this issue, we propose a Robust Self-reflective Hashing (RSH) framework that prudently analyzes semantic discrepancies while accounting for uncertainty, thereby effectively mitigating interference from noisy labels. Specifically, the Double Feature Representation (DFR) method is introduced, employing semantic and uncertainty features to represent the semantic representation and fuzziness of samples. With a double feature, we propose a novel cross-modal similarity metric - the Self-reflective Similarity Metric (SSM), which judges sample similarity by integrating semantic discrepancy and fuzziness, enabling the model to adaptively weaken semantic discrepancy according to uncertainty level. The proposed method is plug-and-play, enabling seamless integration into diverse objective functions to enhance model robustness and reliability. Extensive experiments on benchmark datasets demonstrate that RSH outperforms existing methods. Code is available at https://github.com/QinLab-WFU/RSH.

Large-scale search, recommendation, and retrieval-augmented generation (RAG) systems typically employ a two-stage architecture: an early-stage ranker (ESR) generates a candidate set, which is subsequently re-ranked by a late-stage ranker (LSR). While there are many reinforcement learning (RL) methods for training the LSR, end-to-end training of the ESR has proven challenging. In particular, naive application of "vanilla" policy gradient (V-PG) is not scalable for candidate-set sizes relevant for practical use due to exploding variance. This issue arises because V-PG propagates the gradient to the joint probability of the candidate sets, ignoring the contribution of each specific item in the candidate set to the reward. To mitigate this issue, we propose a novel **"credit-assigned" policy gradient (CA-PG)**, which computes gradients with respect to the probability that the target item is chosen in any candidate set, i.e. marginalizing over all candidate sets that contain it. Our theoretical analysis reveals that CA-PG significantly reduces the variance of V-PG by marginalizing over the specific composition of the candidate set, while preserving the ability to learn the correct ranking of items under a reasonably aligned LSR policy. Experiments on both synthetic and real-world data demonstrate that CA-PG improves the convergence speed and training stability for ESRs utilizing the canonical Plackett-Luce model, especially when the candidate-set size is large.

Recent advancements in tabular deep learning have demonstrated exceptional practical performance, yet the field often lacks a clear understanding of why these techniques actually succeed. To address this gap, our paper highlights the importance of the concept of data (aleatoric) uncertainty for explaining the effectiveness of recent tabular DL methods. While data uncertainty leads to irreducible prediction errors on test samples, it also introduces stochasticity into the training signal that can impede effective learning. We demonstrate that tabular methods differ significantly in their ability to cope with this optimization challenge. Specifically, we reveal that the success of many beneficial design choices in tabular DL, such as numerical feature embeddings, advanced ensembling strategies, retrieval-augmented models, and tabular Prior-Fitted Networks, can be partially attributed to their respective implicit mechanisms for performing well under high data uncertainty. By dissecting these varied mechanisms, we provide a unifying understanding of recent performance improvements. Furthermore, leveraging insights from this perspective, we design a novel, more effective numerical feature embedding method as an immediate practical outcome of our analysis. Overall, our work paves the way toward a principled understanding of the benefits introduced by modern tabular methods that results in the concrete advancements of existing techniques and outlines future research directions for tabular DL.

Search-integrated reasoning enables language agents to transcend static parametric knowledge by actively querying external sources. However, training these agents via reinforcement learning is hindered by the *multi-scale credit assignment* problem: existing methods typically rely on sparse, trajectory-level rewards that fail to distinguish between high-quality reasoning and fortuitous guesses, leading to redundant or misleading search behaviors. To address this, we propose Search-R2, a novel Actor–Refiner collaboration framework that enhances reasoning through targeted intervention, with both components jointly optimized during training. Our approach decomposes the generation process into an Actor, which produces initial reasoning trajectories, and a Meta-Refiner, which selectively diagnoses and repairs flawed steps via a ``cut-and-regenerate'' mechanism. To provide fine-grained supervision, we introduce a hybrid reward design that couples outcome correctness with a dense process reward quantifying the information density of retrieved evidence. Theoretically, we formalize the Actor–Refiner interaction as a smoothed mixture policy, proving that selective correction yields strict performance gains over strong baselines. Extensive experiments across various general and multi-hop QA datasets demonstrate that Search-R2 consistently outperforms strong RAG and RL-based baselines across model scales, achieving superior reasoning accuracy with minimal overhead.

The insatiable demand for web-scale training data has exposed LLMs to a subtle but consequential threat: the absorption of malicious scam content into model weights and its subsequent reproduction during inference. In November 2024, this risk materialized when a developer reportedly lost 2,500 USD after ChatGPT generated an otherwise routine cryptocurrency trading script containing a live phishing URL. To systematically investigate this problem, we introduce Scam2Prompt, an automated auditing framework that crawls known scam websites, infers their functional intent, and synthesizes innocuous developer-style prompts — the kind of legitimate coding requests a programmer might naturally submit — to evaluate whether LLMs reproduce the underlying scam endpoints. Importantly, our approach requires neither jailbreaking nor adversarial prompting; all 1,377 prompts in our benchmark, Innoc2Scam-bench, which is automatically constructed by Scam2Prompt, were human-validated as benign coding tasks. Evaluation of seven production LLMs released in 2025 on Innoc2Scam-bench shows that the vulnerability proves both persistent and severe: malicious code generation rates range from 12.9% to 47.3% across the evaluated models, and no tested model proves immune. State-of-the-art guardrails and RAG-based agents offer only limited protection, underscoring an urgent need for explicit URL validation in LLM-assisted software development pipelines.

Encoder-only language models are frequently used for a variety of language tasks, including classification and retrieval. However, there has been a lack of recent research efforts for encoder models, especially with respect to multilingual models. We introduce mmBERT, an encoder-only language model pretrained on 3T tokens of multilingual text in over 1800 languages. To build mmBERT we introduce several novel elements to massively multilingual encoder training, including phased data curation and scheduled language inclusion. We add over 1700 low-resource languages to the data mix only during the decay phase, showing that it boosts performance dramatically and maximizes the gains from the relatively small amount of training data without excessive repetition. The model uses recent advances in architecture and training schemes to be faster and more multilingual than other models and we release weights, data, and code. We show that mmBERT significantly outperforms the previous generation, on various tasks, for both high and low-resource languages.

Recent Large Language Model based approaches for clarifying visual design largely focus on selecting questions that better uncover user intent, but often overlooks the cognitive burden imposed on users, i.e., the effort required to interpret and answer these questions, which is crucial for effective human-agent interaction. In this paper, we propose Agentic Model Predictive Questioning Control (A-MPQC), a test-time framework that reduces proxy-estimated user interaction burden while improving visual design alignment by formulating multi-round clarification as trajectory optimization with receding-horizon replanning to revise its questioning strategy. In addition, we introduce lookahead question plans to reduce ambiguity early, and a lightweight respond-or-reject surrogate reward to steer questions toward lower user-burden formats (e.g., yes/no). Experiments on webpage and ad banner generation benchmarks show that A-MPQC not only generates designs better aligned with user intent, but also achieves lower user-interaction cost across diverse interaction baselines, including fixed-format strategies (e.g., multiple-choice and open-ended) and a retrieval-augmented baseline, without retraining. This paper sets a new perspective that explicitly formulates and optimizes the human cognitive burden jointly with final design alignment, opening new opportunities to advance human-agent interaction.

Contrastive learning has advanced the representation learning across domains, yet its success relies on data augmentations that preserve semantic contents while providing the view diversities. Multivariate time series, however, are inherently noisy, non-stationary, and lack such intuitive semantic cues. Consequently, standard heuristic augmentations that ignore semantic parts may risk destroying critical temporal dependencies. Though some recent approaches attempt to isolate informative components, they typically rely on an implicit neural mechanism to infer semantics, thus limiting the interpretability and controllability. To address this, we propose ProSAR, an information-theoretic framework that leverages the explicit prototype alignment to guide semantic augmentations, and establish a feedback loop between the augmentation, contrastive learning, and prototype updates. Specifically, grounded in our proposed Prototype-Conditioned Information Bottleneck principle, we leverage the time-domain prototypes as explicit anchors to localize semantic segments, and develop a time–frequency augmentation strategy that retains prototype-consistent information while discarding noise. To promote semantically consistent prototypes for a reliable view generation, we design a dual-prototype loop where the augmented views are encoded into representations and then the learned representations are clustered to update latent prototypes, whose decoded feedback refines the time-domain prototypes for the next round of augmentation. Experiments on diverse time-series benchmarks demonstrate that ProSAR outperforms the other contrastive learning methods on downstream forecasting and classification tasks.

Large Language Models (LLMs) achieve remarkable performance but struggle with complex logical reasoning, particularly in real-world settings. Existing research is largely confined to the closed-world assumption, which posits that all premises required for reasoning are explicitly provided. However, real-world tasks frequently exhibit open-world characteristics, where the provided information is insufficient to infer a conclusion due to missing premises or implicit commonsense knowledge. To address this, we propose OpenIKLR, an Open-world Incomplete-Knowledge-aware Logical Reasoning framework that integrates symbolic logic solvers with LLMs. OpenIKLR first translates natural language into symbolic representations to pinpoint reasoning gaps via a logical solver. It then iteratively generates a minimal set of necessary missing premises using LLMs. To ensure these added premises are both logically sound and factually accurate, we introduce a dual-verification: logic verification via the solver and fact verification via the LLMs. Experiments show that OpenIKLR consistently outperforms existing logical reasoning and RAG baselines across multiple backbones and real-world datasets. The code is available at https://github.com/OpenIKLR/OpenIKLR.

Time series forecasting plays a pivotal role in data-driven decision-making across various time series domains. Recently, leveraging their ability to extract semantically rich representations, Large Language Models (LLMs) have achieved promising results in time series forecasting. However, existing LLM-based methods struggle to obtain multi-scale retrieval-augmented representations due to entangled multi-scale representations and redundant multi-scale interference. To address this, we propose TimeMRA, an LLM-empowered Time series forecasting framework via Multi-Scale Retrieval-Augmented representations. Specifically, a scale-aware prompt generation (SAPG) module is designed to decompose time series into multiple scales and generate augmented multi-scale representations. Then, a cross-scale disentanglement constraint (CSDC) mechanism with a router network is designed to obtain the disentangled multi-scale semantic representations while mitigating interference from irrelevant scales. Finally, a cross-modality retrieval module is designed to obtain multi-scale retrieval-augmented representations for time series forecasting. Experiments on 10 real-world datasets demonstrate that TimeMRA achieves state-of-the-art (SOTA) performance.

Effectively retrieving, reasoning, and understanding multimodal information remains a critical challenge for agentic systems. Traditional Retrieval-augmented Generation (RAG) methods rely on linear interaction histories, which struggle to handle long-context tasks, especially those involving information-sparse yet token-heavy visual data in iterative reasoning scenarios. To bridge this gap, we introduce VimRAG, a framework tailored for multimodal Retrieval-augmented Reasoning across text, images, and videos. Inspired by our systematic study, we model the reasoning process as a dynamic directed acyclic graph that structures the agent states and retrieved multimodal evidence. Building upon this structured memory, we introduce a Graph-Modulated Visual Memory Encoding mechanism, with which the significance of memory nodes is evaluated via their topological position, allowing the model to dynamically allocate high-resolution tokens to pivotal evidence while compressing or discarding trivial clues. To implement this paradigm, we propose a Graph-Guided Policy Optimization strategy. This strategy disentangles step-wise validity from trajectory-level rewards by pruning memory nodes associated with redundant actions, thereby facilitating fine-grained credit assignment. Extensive experiments demonstrate that VimRAG consistently achieves state-of-the-art performance on diverse multimodal RAG benchmarks.

Long-context inference is constrained by KV-cache memory, which grows linearly with sequence length; KV-cache compression therefore hinges on reliably selecting which past tokens to retain. Most geometry-based eviction methods score keys by cosine similarity to a global centroid, but cosine is scale-invariant and can discard magnitude cues that distinguish semantically salient tokens. We propose ManifoldKV, a training-free scorer that ranks tokens by Euclidean distance to the key centroid, capturing both angular and radial deviations. On the RULER benchmark, ManifoldKV achieves **95.7%** accuracy at 4K–16K contexts with 20% compression, matching the best geometric baseline overall while *decisively* outperforming it in two regimes where magnitude information is critical. First, on multi-key retrieval, ManifoldKV reduces directional collisions, achieving **92.4%** vs KeyDiff's 77.0% (+15.4 points) on 3-key NIAH at 50% compression. Second, to address dilution and performance collapse of global centroids at 64K context, we introduce WindowedManifoldKV, which restores accuracy to 84.3% at 25% compression, a 49-point recovery over global L2 and +3.2 points over KeyDiff. Beyond RULER, we validate on real-world benchmarks: on LongBench, ManifoldKV outperforms KeyDiff by +2.80 points on Qwen3-8B (winning 12 of 14 tasks) and +0.49 on Phi-4; on HELMET, ManifoldKV achieves +6.5 EM on RAG and WindowedManifoldKV reaches +42 points on multi-key recall at 131K; and on InfiniteBench at 100K+ context, WindowedManifoldKV wins by +7.16 on Phi-4. Cross-architecture evaluation across six models reveals that the optimal distance metric depends on key-norm geometry, providing the first systematic guidelines for metric selection in geometric KV cache compression. The method requires only 3 lines of code and works across diverse architectures without tuning.

With the rapid emergence of personal AI agents based on Large Language Models (LLMs), implementing them on-device has become essential for privacy and responsiveness. To handle the inherently personal and context-dependent nature of real-world requests, such agents must ground their generation in device-resident personal context. However, under tight memory budgets, the core bottleneck is *what to store* so that retrieval remains aligned with the user. We propose EPIC (Efficient Preference-aligned Index Construction), which focuses on user preferences as a compact and stable form of personal context and integrates them throughout the RAG pipeline. EPIC selectively retains preference-relevant information from raw data and aligns retrieval toward preference-aligned contexts. Across four benchmarks covering conversations, debates, explanations, and recommendations, EPIC reduces indexing memory by 2,404$\times$, improves preference-following accuracy by 18.79\%p, and achieves 32.17$\times$ lower retrieval latency over the best-performing baseline. In on-device experiments, EPIC maintains under 1 MB memory and achieves 5.21 to 29.35 ms/query latency across three platforms, while supporting streaming updates under preference drift. Our code and data are available at https://github.com/UbiquitousAILab/EPIC.

While RAG systems are increasingly deployed in dynamic web search, temporal volatility amplifies their vulnerability to adversarial attacks. Existing static-oriented defenses struggle to handle evolving threats and incur prohibitive storage costs in dynamic settings. We propose RADAR, a framework that models reliable context selection as a graph-based energy minimization problem, solved exactly via Max-Flow Min-Cut. By incorporating a Bayesian memory node, RADAR recursively updates a belief state instead of archiving raw historical documents, effectively balancing stability against attacks with adaptability to genuine knowledge shifts. Experiments on a novel dynamic dataset show that RADAR achieves superior robustness and response quality with minimal storage overhead compared to the baselines.