ICML 2026 Papers

Inference from quantized data has received significant attention in recent years due to its broad applications in machine learning and signal processing. Existing likelihood-based approaches are often restricted to Gaussian assumptions or low-bit quantization settings, limiting modeling flexibility and robustness under complex data distributions. In this work, we study inference from quantized observations under the general normal variance-mean mixture (NVMM) framework, which encompasses distributions including Gaussian, $t$, generalized hyperbolic skew-$t$, and generalized hyperbolic distributions. Optimization under the NVMM framework is challenging because the underlying likelihood function involves multidimensional integrals that are difficult to evaluate due to multidimensional quantization and latent mixture variables. To address this difficulty, we propose an expectation conditional maximization (ECM) algorithm with latent-variable augmentations for both quantization and mixture modeling. By leveraging the conditional Gaussian structure of the NVMM family, the proposed method admits closed-form updates for all model parameters at each iteration, leading to an efficient and tractable optimization procedure. We further establish global linear convergence guarantees for the proposed ECM algorithm. Beyond basic parameter estimation, the proposed framework naturally extends to several structured learning and recovery tasks under the NVMM framework, including quantized regression, matrix completion, compressive sensing, and covariance estimation. Numerical experiments demonstrate the effectiveness and robustness of the proposed framework across a variety of quantized inference problems.

Clustering is commonly compared through leaderboards that collapse performance into a single aggregate ranking. Such summaries obscure why methods succeed, which data properties align with failure, and how conclusions shift under representation changes and realistic tuning constraints. We present CHB, a diagnostic toolkit for hardness-aware clustering evaluation. CHB maps each dataset--representation pair to an interpretable hardness fingerprint capturing (i) separation, (ii) cohesion and scale heterogeneity, and (iii) topology through scalable persistent-homology summaries. Using this diagnostic space, CHB evaluates clustering algorithms under standardized, compute-aware tracks. Conditioning results on hardness coordinates turns comparison into diagnosis: across a broad range of datasets and their representations, CHB reveals reproducible structural regimes, uncovers regime-dependent ranking reversals across method families, and surfaces robustness signatures, including topology-linked breakdowns. CHB further enables representation auditing by attributing gains to measurable shifts in the hardness fingerprint rather than just external performance changes. We release CHB as an open, extensible artifact for evaluating new clustering methods and embeddings within a shared diagnostic framework.

Clustering in a big data setting is an intensively studied problem, with coresets emerging as one of the important paradigms in this line of work. Given a cost function $\text{cost}(P,S)$ mapping input points $P$ and a solution $S$ to an objective value, a coreset is a typically weighted sketch $\Omega\subseteq P$ such that $\text{cost}(\Omega,S)\approx \text{cost}(P,S)$. In practice, coreset sizes much smaller than those suggested by theoretical guarantees are often found to be sufficient. In this paper, we offer an explanation for this phenomenon. Smaller coreset sizes suffice if we only wish to preserve the costs of \emph{good} solutions, i.e., solutions with low cost. We define and devise \emph{approximation-preserving coresets}, which provide a weaker guarantee than strong coresets, which apply to all solutions, while providing stronger guarantees than weak coresets, which apply only to the optimum solution. We complement this result by showing that even a very small distortion in the approximation factor cannot admit coresets of this size.

Hybrid sequence models—combining Transformer and state-space model layers—seek to gain the expressive versatility of attention as well as the computational efficiency of state-space model layers. Despite burgeoning interest in hybrid models, we lack a basic understanding of the settings where—and underlying mechanisms through which—they offer benefits over their constituent models. In this paper, we study this question, focusing on a broad family of core synthetic tasks. For this family of tasks, we prove the existence of fundamental limitations for non-hybrid models. Specifically, any Transformer or state-space model that solves the underlying task requires either a large number of parameters or a large working memory. On the other hand, for two prototypical tasks within this family—namely selective copying and associative recall—we construct hybrid models of small size and working memory that provably solve these tasks, thus achieving the best of both worlds. Our experimental evaluation empirically validates our theoretical findings. Importantly, going beyond the settings in our theoretical analysis, we empirically show that learned—rather than constructed—hybrids outperform non-hybrid models with up to $6 \times$ as many parameters. We additionally demonstrate that hybrid models exhibit stronger length generalization and out-of-distribution robustness than non-hybrids.

Online evaluation of large language models increasingly relies on sequentially collected pairwise preferences, enabling human-aligned assessment and continuous data collection until closely performing models can be reliably distinguished. However, adaptive sampling and continuous monitoring invalidate classical fixed-sample inference, rendering existing ranking procedures largely heuristic. We propose SERPANT (Sequential E-value Ranking and Pruning via Adaptive Null Testing), a principled framework for online LLM ranking with anytime-valid guarantees. SERPANT formulates model comparison as a collection of pairwise hypothesis tests and constructs e-processes to ensure family-wise error rate control at any monitoring time. Anytime validity provides a theoretical justification for early stopping, enabling substantial cost savings from expensive human annotation. To improve efficiency, we introduce a novel tournament-based sampling strategy that adaptively selects comparisons based on past outcomes. The proposed framework is also extended to top-k model identification.

Multi-agent LLM debates are increasingly used in domains such as policy, politics, and city planning, where ground truth is often unavailable. Yet existing evaluations rely heavily on outcome-based proxies such as consensus, majority vote, or LLM-as-judge scores, which can miss failures like sycophancy, domination, and premature convergence. We introduce a diagnostic framework that evaluates both debate outcomes and the deliberative process using interpretable metrics for engagement, responsiveness, influence asymmetry, balance, stability, and agent utility. Across real-world debate settings and validation benchmarks, our process-level diagnostics align more closely with human judgments and reveal interaction failures that standard outcome-only measures overlook. These results show that reliable evaluation of multi-agent debates requires measuring not only what answer agents reach, but how they reach it.

Neural networks trained under different hyperparameter settings can fall into distinct training "regimes," with consistent behavior within regimes and qualitative differences across regimes. In this paper, we study such multi-regime behavior in scientific machine learning (SciML) models through a *regime-aware diagnostic framework* that jointly analyzes performance, training dynamics, and loss-landscape geometry. We identify three key findings: (i) a consistent three-regime structure emerges across many standard SciML models, different constraint enforcements, and various optimizer designs; (ii) optimization effectiveness is regime-specific, with no single method performing well across all regimes; and (iii) SciML models can exhibit fine-grained failure modes that can challenge conventional interpretations of standard loss-landscape metrics. Our results provide an approach to establish a unified, task-oblivious perspective on failure modes in SciML and to inform regime-aware guidance for improving robustness. We validate these findings across widely-used SciML models, including physics-informed neural networks, neural operators, and neural ordinary differential equations, on benchmarks spanning representative ordinary and partial differential equations.

Large Language Models (LLMs) have demonstrated sophisticated comprehension of sequential code, yet their capacity for reasoning about concurrent programs remains largely unquantified. We introduce DRPBench, a benchmark designed to evaluate the concurrent code comprehension of LLMs by measuring their data race prediction performance. To address the challenge of runtime non-determinism for evaluation on concurrent programs, we frame the evaluation as a fine-grained static prediction task using 1,003 programs from the SV-COMP suite, featuring 557 manually annotated data races with precise variable- and line-level granularity. Our evaluation of 15 state-of-the-art LLMs—spanning standard, reasoning, and agentic variants—reveals that DRPBench effectively differentiates concurrent code comprehension capabilities of LLMs. While the top-performing model (Gemini 3 with test-time reasoning) achieves an F1 score of 74.89%, most models struggle significantly (scoring less than 60%), with Llama 3 70B achieving only 8.80%. Beyond benchmarking, we characterize two primary failure modes: (1) shared-variable distraction, where multiple variable appearances degrade comprehension accuracy, and (2) synchronization-logic myopia, the inability to interpret non-standard synchronization implementations. Our findings provide a diagnostic roadmap for enhancing concurrent code comprehension of LLMs in future development.

Multimodal large language models (MLLMs) are trained on massive multimodal data, making data unlearning increasingly important as data owners may request the removal of specific content. In practice, these requests often arrive sequentially over time, giving rise to the challenging problem of *MLLM Lifelong Unlearning*. However, most existing benchmarks are limited in scale and scope, failing to capture the complexities of MLLM lifelong unlearning. To fill this gap, we introduce the MLUBench, a large-scale and comprehensive benchmark featuring 127 entities across 9 classes under lifelong unlearning requests. We perform extensive experiments using MLUBench and reveal that existing unlearning methods suffer from severe, cumulative degradation. More critically, we further identify the unique challenge of this problem: unlike in unimodal models, MLLM lifelong unlearning is constrained by the need to preserve multimodal alignment. Continually unlearning from one modality could degrade the entire model. To alleviate this challenge, we propose LUMoE, an effective and efficient method. Experiments demonstrate that LUMoE significantly mitigates the degradation problem faced by baselines. The source code and the MLUBench dataset are open-sourced in this [URL](https://github.com/lihe-maxsize/Lifelong_Unlearning_main).

Literature reveals that a Large Language Model's (LLM) behavior is not only conditioned by its original weights but also its instance-level parameters, such as instructional prompt, sampling configuration or quantization. A model that generates safe outputs under one configuration may produce toxic content under another. However, current LLM identification techniques (such as fingerprinting) focus on intellectual property protection, and their design favors robustness to changes in these instance-level parameters. This poses a critical challenge for AI regulation in which compliance assessments target actual deployed behaviors, not model provenance. In this paper, we introduce instance-level fingerprinting, a regulator-oriented paradigm that distinguishes configurations of the same LLM. Our method FLIPS, exploits biases in generated binary random sequences to reach 96% (closed-set) and 90% (open-set, where some targets are unknown) identification accuracy across 237 model instances, versus 35% for the adapted LLMmap baseline. This shows that instance-level fingerprinting is both necessary for regulation and practically feasible. Code available at https://github.com/GurvanR/FLIPS-LLM-Instance-Fingerprinting.

Multi-model learning has attracted great attention in visual-text tasks. However, visual-tabular data, which plays a pivotal role in high-stakes domains like healthcare and industry, remains underexplored. In this paper, we introduce \textit{VT-Bench}, the first unified benchmark for standardizing vision-tabular discriminative prediction and generative reasoning tasks. VT-Bench aggregates 14 datasets across 9 domains (medical-centric, while covering pets, media, and transportation) with over 756K samples. We evaluate 23 representative models, including unimodal experts, specialized visual-tabular models, general-purpose vision-language models (VLMs), and tool-augmented methods, highlighting substantial challenges of visual-tabular learning. We believe VT-Bench will stimulate the community to build more powerful multi-modal vision-tabular foundation models. Benchmark: \url{https://github.com/LAMDA-NeSy/VT-Bench}

Evaluation of long-context Large Language Models (LLMs) has advanced rapidly. However, most existing benchmarks are limited to the document level and focus mainly on high-resource languages, leaving many fine-grained challenges insufficiently evaluated. To address this gap, we present MGAL, the first multilingual, granularity- and position-aware long-context benchmark. MGAL is constructed from United Nations (UN) reports spanning 8K to 128K tokens across the six official UN languages. It covers four coherent levels of linguistic granularity (word, sentence, paragraph, and document) and further stratifies entries by their position within the document (begin, middle, and end), indexed at both the document and paragraph levels. This design enables systematic diagnosis of multilingual long-context comprehension across different granularities. Through extensive experiments and analyses, we find that: (1) LLMs perform well at word-level tasks but struggle with coarser-grained ones; and (2) Closed-source models retain a clear performance advantage in lower-resource languages. We further identify two new challenges: (1) Under local semantic crowding, where neighboring sentences share topics and entities, models tend to follow surface cues (e.g., connectives like 'however' or repeated entities) rather than the discourse role of the sentence in surrounding context (e.g., background, outcome); and (2) A gap between fluency and consistency in generated outputs, where models produce text that reads smoothly but drifts from the source facts. In addition, we observe several patterns in line with prior studies, including reliance on nearby evidence and reuse of options under uncertainty.

AI agents are increasingly deployed to execute important tasks. While rising accuracy scores on standard benchmarks suggest rapid progress, many agents still continue to fail in practice. This discrepancy highlights a major limitation of current evaluations: focusing on a single metric is not enough to understand agent behavior. Notably, it ignores whether agents behave consistently across runs, withstand perturbations, fail predictably, or have bounded error severity. Grounded in safety-critical engineering, we provide a holistic performance profile consisting of twelve metrics that decompose agent reliability along four key dimensions: *consistency*, *robustness*, *predictability*, and *safety*. Evaluating 15 models across two complementary benchmarks, we find that recent capability gains have only yielded small improvements in reliability. By exposing these persistent limitations, our metrics complement traditional evaluations while offering tools for reasoning about how agents perform, degrade, and fail.

Conditional generative models have recently achieved remarkable success in various applications. However, a suitable metric for evaluating the reliability of these models, which takes into account their inherent uncertainty, is still lacking. Existing metrics, which typically assess a single output, may fail to capture the variability or potential risks in generation. In this paper, we propose a novel evaluation metric called reliability score based on conformal prediction, which measures the worst-case performance within the prediction set at a pre-specified confidence level. However, computing this score is challenging due to the high-dimensional nature of the output space and the nonconvexity of both the metric function and the prediction set. To efficiently compute this score, we introduce Conformal ReLiability (CReL), a framework that can (i) construct the prediction set with desired coverage; and (ii) accurately optimize the reliability score within the constructed prediction set. We provide theoretical results on coverage and demonstrate empirically that our method produces more informative prediction sets than existing approaches. Experiments on synthetic data and the image-to-text and text-to-image tasks further demonstrate the interpretability of our new metric, and the validity and effectiveness of our computational framework. Source code can be found at https://ggc29.github.io/CReL/.

Distribution shift poses a fundamental challenge in offline reinforcement learning, often leading to value overestimation when querying out-of-distribution actions. We introduce Regularized Evaluation Gap (REG) as a bridge between implicit methods like IQL and explicit conservative methods. We formulate policy evaluation as a robust optimization problem over an ambiguity set of critics and show that IQL’s objective can be viewed as an approximate dual solution to this problem. To extract a policy from the learned value function, we propose a practical Orthogonal Policy Gradient (OPG) update. This method regularizes an aggressive, mode-seeking policy gradient by projecting it onto the subspace orthogonal to a stable, in-sample behavior cloning gradient. Extensive D4RL experiments demonstrate that REG matches state-of-the-art performance among both Gaussian methods and diffusion-based approaches without the computational burden of the latter.

Vision-language models (VLMs) generate fluent causal explanations, but current evaluations cannot distinguish linguistic plausibility from faithful causal reasoning. We introduce a dual-probe methodology that isolates these properties. The Text-Only Probe measures linguistic quality. The Chain-Text Probe requires models to first generate explicit causal chains. The Abstraction Gap (AG) metric quantifies the normalized performance difference. Evaluating eight VLMs on CAGE (Causal Abstraction Gap Evaluation), a benchmark of 49,500 questions across 5,500 images spanning Pearl's causal hierarchy, we find seven models exhibit AG exceeding 0.50 with text scores of 6--8 but chain scores below 2.5. Fine-tuning on 45,000 chain-annotated examples fails to close the gap. However, one model achieves near-zero AG. The capability exists within current VLM architectures and depends on pretraining and architectural choices. CAGE provides a diagnostic tool for assessing faithful causal reasoning in VLMs.

Large language model (LLM)-driven multi-agent systems typically require multiple model invocations and complex coordination during inference, and their execution strategies directly affect system accuracy, latency, and computational cost. Parallel execution provides a means to improve inference-time efficiency. From the perspective of inference-time execution, this paper models parallelism in multi-agent systems as two distinct levels of decision processes: Replica Parallelism, which explores multiple complete solution paths at the task level, and Structural Parallelism, which enables concurrent execution within a single solution path through task decomposition. However, the roles of different forms of parallelism and their interrelationships still lack systematic study in terms of unified organization and coordination. We therefore propose TIPEX, a controllable execution framework that unifies these two levels of parallelism and coordinates their roles within the inference process under a unified execution semantics while supporting systematic combinations and analyses of different parallel strategies and parameter configurations. Systematic experiments on the GAIA benchmark demonstrate that inference-time parallelism can significantly improve accuracy and reduce end-to-end latency at the cost of increased token consumption. Further analysis shows that Replica and Structural Parallelism exhibit complementary effects across task complexities, with tasks of intermediate difficulty benefiting most from their coordination, while overly aggressive parallel strategies do not necessarily yield better performance.

Data science agents promise to accelerate discovery and insight-generation by turning data into executable analyses and findings. Yet existing data science benchmarks fall short due to fragmented evaluation interfaces that make cross-benchmark comparison difficult, narrow task coverage and a lack of rigorous data grounding. In particular, we show that a substantial portion of tasks in current benchmarks can be solved without using the actual data. To address these limitations, we introduce DSGym, a standardized framework for evaluating and training data science agents in self-contained execution environments. Unlike static benchmarks, DSGym provides a modular architecture that makes it easy to add tasks, agent scaffolds, and tools, positioning it as a live, extensible testbed. We curate DSGym-Tasks, a holistic task suite that standardizes and refines existing benchmarks via quality and shortcut solvability filtering. We further expand coverage with (1) DSBio: expert-derived bioinformatics tasks grounded in literature and (2) DSPredict: challenging prediction tasks spanning domains such as computer vision, molecular prediction, and single-cell perturbation. Beyond evaluation, DSGym enables agent training via execution-verified data synthesis pipeline. As a case study, we build a 2,000-example training set in DSGym that substantially improves a 4B mode on standardized analysis benchmarks. Overall, DSGym enables rigorous end-to-end measurement of whether agents can plan, implement, and validate data analyses in realistic scientific context.

Reliable evaluation of large language model (LLM) agents depends critically on benchmark validity, yet modern agent benchmarks often contain hidden flaws arising from interactions among user instructions, environments, tools, ground-truth trajectories, and evaluation protocols. These flaws confound model errors with benchmark artifacts and undermine leaderboard-based comparisons. We propose **COBA** (**CO**mponent-based **B**enchmark **A**uditing), an automated pipeline for diagnosing and filtering validity issues in agent benchmarks. COBA decomposes agent tasks into four standardized components—User, Environment, Ground Truth, and Evaluation—and operationalizes a component-level issue taxonomy using hybrid rule-based detectors and taxonomy-guided LLM evaluation. Across six widely used agent benchmarks, COBA achieves strong alignment with expert judgments, with F1 scores between 0.791 and 0.874. It complements manual verification of $\tau^2$-bench by identifying issues missed due to benchmark complexity, and generalizes to previously unseen benchmarks with minimal adaptation. Our analysis shows that benchmark flaws are widespread and materially affect evaluation outcomes, demonstrating that component-based auditing provides a scalable foundation for more reliable and interpretable agent evaluation. We release AgentSuite, a unified benchmark-running platform that includes the COBA auditing pipeline and audited benchmark variants: https://github.com/Agent-Suite/AgentSuite.