ICLR 2026 论文列表

Diffusion Language Models (DLMs) present a compelling alternative to autoregressive models, offering flexible, any-order infilling without specialized prompting design. However, their practical utility is blocked by a critical limitation: the requirement of a fixed-length masked sequence for generation. This constraint severely degrades code infilling performance when the predefined mask size mismatches the ideal completion length. To address this, we propose DreamOn, a novel diffusion framework that enables dynamic, variable-length generation. DreamOn augments the diffusion process with two length control states, allowing the model to autonomously expand or contract the output length based solely on its own predictions. We integrate this mechanism into existing DLMs with minimal modifications to the training objective and no architectural changes. Built upon Dream-Coder-7B and DiffuCoder-7B, DreamOn achieves infilling performance on par with state-of-the-art autoregressive models on HumanEval-Infilling and SantaCoder-FIM and matches oracle performance achieved with ground-truth length. Our work removes a fundamental barrier to the practical deployment of DLMs, significantly advancing their flexibility and applicability for variable-length generation. Our code is available at https://github.com/DreamLM/DreamOn.

Generating a street-level 3D scene from a single satellite image is a crucial yet challenging task. Current methods present a stark trade-off: geometry-colorization models achieve high geometric fidelity but are typically building-focused and lack semantic diversity. In contrast, proxy-based models use feed-forward image-to-3D frameworks to generate holistic scenes by jointly learning geometry and texture, a process that yields rich content but coarse and unstable geometry. We attribute these geometric failures to the extreme viewpoint gap and sparse, inconsistent supervision inherent in satellite-to-street data. We introduce Sat3DGen to address these fundamental challenges, which embodies a geometry-first methodology. This methodology enhances the feed-forward paradigm by integrating novel geometric constraints with a perspective-view training strategy, explicitly countering the primary sources of geometric error. This geometry-centric strategy yields a dramatic leap in both 3D accuracy and photorealism. For validation, we first constructed a new benchmark by pairing the VIGOR-OOD test set with high-resolution DSM data. On this benchmark, our method improves geometric RMSE from 6.76m to 5.20m. Crucially, this geometric leap also boosts photorealism, reducing the Fr\'echet Inception Distance (FID) from $\sim$40 to 19 against the leading method, Sat2Density++, despite using no extra tailored image-quality modules. We demonstrate the versatility of our high-quality 3D assets through diverse downstream applications, including semantic-map-to-3D synthesis, multi-camera video generation, large-scale meshing, and unsupervised single-image Digital Surface Model (DSM) estimation.

Aligning Large Language Models (LLMs) with human preferences through finetuning is resource-intensive, motivating lightweight alternatives at test time. We address test-time alignment through the lens of sequential decision making, a perspective that reveals two fundamental challenges. When actions are defined at the token level, as in guided decoding, alignment suffers from the curse of horizon. Conversely, when actions are at the response level, as in traditional iterative refinement, the curse of dimensionality emerges. To resolve this trade-off, we draw inspiration from Model Predictive Control (MPC) in control theory to propose Textual Model Predictive Control (TMPC), a novel predictive planning framework adapted for aligning LLMs at inference time. A key limitation of standard MPC is its reliance on predefined, hard segment boundaries, which are often absent in text generation. TMPC overcomes this by introducing two principles inspired by hierarchical reinforcement learning: (1) Hindsight Subgoal Identification, where TMPC analyzes generation subgoals to retrospectively identify high-reward intermediate outputs as subgoals. This allows the framework to discover meaningful, task-specific planning steps (e.g., a sentence in machine translation or a bug fix in code generation.). (2) Subgoal-Conditioned Re-Generation, where these identified subgoals are used to guide subsequent planning iterations. By conditioning on these proven, high-quality subgoals, TMPC ensures stable improvement by building upon previously validated successes. TMPC is evaluated on three tasks with distinct segmentation properties: discourse-level translation, long-form response generation, and program synthesis. The results demonstrate that TMPC consistently improves performance, highlighting the generality.

Reinforcement Learning from Verifiable Rewards (RLVR) has significantly improved the performance of large language models (LLMs) on tasks with gold ground truth, such as code generation and mathematical reasoning. However, its application to open-ended question answering (QA) remains challenging, primarily due to the absence of reliable evaluation and verifiable reward signals. This difficulty is further compounded by the limitations of existing evaluation paradigms. Previous approaches typically rely on human feedback or LLM-as-judge strategies, which are costly, prone to reward hacking, and often fail to provide sufficiently discriminative or interpretable evaluation signals. To address these limitations, we introduce a schema for generating case-wise rubrics that are question-specific, content-based and stylistically sensitive, thereby evaluating both factual soundness and writing quality. Building on this schema, we propose QuRL (Open-Ended QA with Rubric-guided Reinforcement Learning), a framework that automatically mines rubrics for each question from easily accessible online sources and leverages them as reward signals. With these rubrics, QuRL employs the GRPO (Group Relative Policy Optimization) algorithm to guide the model in exploring the correct generation path. Extensive experiments show that our framework achieves significant improvements of total +17.0 points on evaluation benchmark, demonstrating the effectiveness of rubric-guided reinforcement learning for open-ended QA.

High computational costs and slow inference hinder the practical application of video generation models. While prior works accelerate the generation process through feature caching, they often suffer from notable quality degradation. In this work, we reveal that this issue arises from their inability to distinguish truly redundant features, which leads to the unintended skipping of computations on important features. To address this, we propose \textbf{PreciseCache}, a plug-and-play framework that precisely detects and skips truly redundant computations, thereby accelerating inference without sacrificing quality. Specifically, PreciseCache contains two components: LFCache for step-wise caching and BlockCache for block-wise caching. For LFCache, we compute the Low-Frequency Difference (LFD) between the prediction features of the current step and those from the previous cached step. Empirically, we observe that LFD serves as an effective measure of step-wise redundancy, accurately detecting highly redundant steps whose computation can be skipped through reusing cached features. To further accelerate generation within each non-skipped step, we propose BlockCache, which precisely detects and skips redundant computations at the block level within the network. Extensive experiments on various backbones demonstrate the effectiveness of our PreciseCache, which achieves an average of $2.6\times$ speedup without noticeable quality loss. Source code will be released.

Logic synthesis (LS), which aims to generate a *compact* logic circuit graph with minimized size while *accurately* satisfying a given functionality, plays an important role in chip design. However, existing LS methods struggle to balance circuit structure compactness and functional accuracy, often leading to suboptimal generation. To address this problem, we propose a novel *Circuit Program Evolution* framework, namely CircuitEvo, which iteratively leverages large language models (LLMs) to evolve circuit programs towards improved compactness while preserving functional accuracy. Specifically, CircuitEvo models the circuit graph as a structured program and leverages the strong generative capabilities of LLMs — guided by domain-specific evolutionary prompt strategies — to generate promising circuit candidates in each iteration. Moreover, a structure-aware circuit optimization module is introduced to correct functional discrepancies by appending necessary substructures to the generated circuits. To the best of our knowledge, CircuitEvo is *the first* LLM-based LS approach that can iteratively improve a circuit's compactness while ensuring functional accuracy. Experiments on several widely used benchmarks demonstrate that CircuitEvo can efficiently generate accurate circuits with up to 16 input number and 69 output number. Moreover, our method significantly outperforms state-of-the-art methods in terms of circuit size, achieving an average improvement of 6.74%.

Large Language Models (LLMs) demonstrate potentials for automating scientific code generation but face challenges in reliability, error propagation in multi-agent workflows, and evaluation in domains with ill-defined success metrics. We present a Bayesian adversarial multi-agent framework specifically designed for AI for Science (AI4S) tasks in the form of a Low-code Platform (LCP). Three LLM-based agents are coordinated under the Bayesian framework: a Task Manager that structures user inputs into actionable plans and adaptive test cases, a Code Generator that produces candidate solutions, and an Evaluator providing comprehensive feedback. The framework employs an adversarial loop where the Task Manager iteratively refines test cases to challenge the Code Generator, while prompt distributions are dynamically updated using Bayesian principles by integrating code quality metrics: functional correctness, structural alignment, and static analysis. This co-optimization of tests and code reduces dependence on LLM reliability and addresses evaluation uncertainty inherent to scientific tasks. LCP also streamlines human-AI collaboration by translating non-expert prompts into domain-specific requirements, bypassing the need for manual prompt engineering by practitioners without coding backgrounds. Benchmark evaluations demonstrate LCP’s effectiveness in generating robust code while minimizing error propagation. The proposed platform is also tested on an Earth Science cross-disciplinary task and demonstrates strong reliability, outperforming competing models.

We introduce AgentSynth, a scalable and cost-efficient pipeline for automatically synthesizing high-quality tasks and trajectory datasets for generalist computer-use agents. Leveraging information asymmetry, AgentSynth constructs subtasks that are simple during generation but significantly more challenging when composed into long-horizon tasks, enabling the creation of over 6,000 diverse and realistic tasks. A key strength of AgentSynth is its ability to precisely modulate task complexity by varying the number of subtasks. Empirical evaluations show that state-of-the-art LLM agents suffer a steep performance drop, from 18\% success at difficulty level 1 to just 4\% at level 6, highlighting the benchmark's difficulty and discriminative power. Moreover, our pipeline achieves a low average cost of \$0.60 per trajectory, orders of magnitude cheaper than human annotations. Code is available in the supplementary materials.

Front-end development constitutes a substantial portion of software engineering, yet converting design mockups into production-ready *User Interface* (UI) code remains tedious and time-costly. While recent work has explored automating this process with *Multimodal Large Language Models* (MLLMs), existing approaches typically rely solely on design images. As a result, they must infer complex UI details from images alone, often leading to degraded results. In real-world development workflows, however, design mockups are usually delivered as Figma files—a widely used tool for front-end design—that embed rich multimodal information (e.g., metadata and assets) essential for generating high-quality UI. To bridge this gap, we introduce Figma2Code, a new task that generalizes *design-to-code* into a multimodal setting and aims to automate *design-to-code* in the wild. Specifically, we collect paired design images and their corresponding metadata files from the Figma community. We then apply a series of processing operations, including rule-based filtering, human and MLLM-based annotation and screening, and metadata refinement. This process yields 3,055 samples, from which designers curate a balanced dataset of 213 high-quality cases. Using this dataset, we benchmark ten state-of-the-art open-source and proprietary MLLMs. Our results show that while proprietary models achieve superior visual fidelity, they remain limited in layout responsiveness and code maintainability. Further experiments across modalities and ablation studies corroborate this limitation, partly due to models’ tendency to directly map primitive visual attributes from Figma metadata.

Accurate auto-formalization of theorem statements is essential for advancing automated discovery and verification of research-level mathematics, yet remains a major bottleneck for LLMs due to hallucinations, semantic mismatches, and their inability to synthesize new definitions. To tackle these issues, we present Aria (**A**gent for **R**etrieval and **I**terative **A**utoformalization), a system for conjecture-level formalization in Lean that emulates human expert reasoning via a two-phase Graph-of-Thought process: recursively decomposing statements into a dependency graph and then constructing formalizations from grounded concepts. To ensure semantic correctness, we introduce **AriaScorer**, a checker that retrieves definitions from Mathlib for term-level grounding, enabling rigorous and reliable verification. We evaluate Aria on diverse benchmarks. On ProofNet, it achieves 91.6\% compilation success rate and 68.5\% final accuracy, surpassing previous methods. On FATE-X, a suite of challenging algebra problems from research literature, it outperforms the best baseline with 44.0\% vs. 24.0\% final accuracy. On a dataset of homological conjectures, Aria reaches 42.9\% final accuracy while all other models score 0\%.

It remains an open question whether LLMs can acquire or generalize genuinely new reasoning strategies, beyond the sharpened skills encoded in their parameters during pre-training or post-training. To attempt to answer this debate, we introduce DELTA-Code —Distributional Evaluation of Learnability and Transferrability in Algorithmic Coding, a controlled benchmark of synthetic coding problem families designed to probe two fundamental aspects: learnability—can LLMs, through reinforcement learning (RL), solve problem families where pretrained models exhibit failure with large enough attempts (pass@K=0)?—and transferrability— if learnability happens, can such skills transfer systematically to out-of-distribution (OOD) test sets? Unlike prior public coding datasets, DELTA isolates reasoning skills through templated problem generators and introduces fully OOD problem families that demand novel strategies rather than tool invocation or memorized patterns. Our experiments reveal a striking grokking phase transition: after an extended period with near-zero reward, RL-trained models abruptly climb to near-perfect accuracy. To enable learnability on previously unsolvable problem families, we explore key training ingredients such as staged warm-up with dense rewards, experience replay, curriculum training, and verification-in-the-loop. Beyond learnability, we use DELTA to evaluate transferability or generalization along exploratory, compositional, and transformative axes, as well as cross-family transfer. Results show solid gains within families and for recomposed skills, but persistent weaknesses in transformative cases. DELTA thus offers a clean testbed for probing the limits of RL-driven reasoning and for understanding how models can move beyond existing priors to acquire new algorithmic skills.

Large Language Models (LLMs) have impressive multilingual capabilities, but they suffer from unexpected code-switching, also known as language mixing, which involves switching to unexpected languages in the model response. This problem leads to poor readability and degrades the usability of model responses. However, existing work on this issue lacks a mechanistic analysis and shows limited effectiveness. In this paper, we first provide an in-depth analysis of unexpected code-switching using sparse autoencoders and find that when LLMs switch to a language, the features of that language exhibit excessive pre-activation values. Based on our findings, we propose $\textbf{S}$parse $\textbf{A}$utoencoder-guided $\textbf{S}$upervised $\textbf{F}$ine$\textbf{t}$uning (SASFT), which teaches LLMs to maintain appropriate pre-activation values of specific language features during training. Experiments on five models across three languages demonstrate that SASFT consistently reduces unexpected code-switching by more than 50\% compared to standard supervised fine-tuning, with complete elimination in four cases. Moreover, SASFT maintains or even improves the models' performance on six multilingual benchmarks, showing its effectiveness in addressing code-switching while preserving multilingual capabilities. The code and data are available at https://anonymous.4open.science/r/SASFT-71CC.

Complex vehicle routing problems (VRPs) remain a fundamental challenge, demanding substantial expert effort for intent interpretation and algorithm design. While large language models (LLMs) offer a promising path toward automation, current approaches still rely on external intervention, which restrict autonomy and often lead to execution errors and low solution feasibility. To address these challenges, we propose an Agentic Framework with LLMs (AFL) for solving complex vehicle routing problems, achieving full automation from problem instance to solution. AFL directly extracts knowledge from raw inputs and enables self-contained code generation without handcrafted modules or external solvers. To improve trustworthiness, AFL decomposes the overall pipeline into three manageable subtasks and employs four specialized agents whose coordinated interactions enforce cross-functional consistency and logical soundness. Extensive experiments on 20 complex VRPs, ranging from standard benchmarks to practical variants, validate the effectiveness and generality of our framework, showing comparable performance against meticulously designed algorithms. Notably, it substantially outperforms existing LLM-based baselines in both code reliability and solution feasibility, achieving rates close to 100% on the evaluated benchmarks.

Autoregressive (AR) models have emerged as a powerful framework for image generation, yet they remain bound by a fundamental limitation: once a prediction is made, it cannot be revised. Each step marches forward in a strict left-to-right sequence, causing small errors to accumulate and compromise the final image. In this work, we reimagine this process with TensorAR, a decoder-only AR model that shifts from predicting discrete tokens to predicting overlapping tensor windows. This simple change transforms image synthesis into a process of next-tensor prediction, enabling the model to refine earlier outputs while preserving the causal structure that defines autoregression. To guard against information leakage during training, we introduce a discrete tensor noising mechanism inspired by discrete diffusion theory, which injects categorical noise into input tensors. TensorAR is designed to be plug-and-play: unlike masked AR methods, it requires no architectural modifications, and unlike autoregressive diffusion, it preserves the familiar AR training paradigm. We evaluate TensorAR across both class-to-image and text-to-image tasks, showing consistent gains in generation quality and instruction-following ability, while achieving a superior balance between quality and latency. In doing so, TensorAR offers a new path forward for autoregressive generation---one where predictions are not just produced, but continually refined.

Autoregressive transformers are the backbone of modern large language models. Despite their success, inference remains slow due to strictly sequential prediction. Prior attempts to predict multiple tokens per step typically impose independence assumptions across tokens, which limits their ability to match the full expressiveness of standard autoregressive models. In this work, we break this paradigm by proposing an efficient and universal framework to jointly predict multiple tokens in a single transformer call, without limiting the representational power. Inspired by ideas from inverse autoregressive normalizing flows, we convert a series of random variables deterministically into a token sequence, incorporating the sampling procedure into a trained model. This allows us to train parallelized models both from scratch and by distilling an existing autoregressive model. Empirically, our distilled model matches its teacher's output for an average of close to 50 tokens on toy data and 5 tokens on a coding dataset, all within a single forward pass.

Recent progress in multimodal generation has increasingly combined autoregressive (AR) and diffusion-based approaches, leveraging their complementary strengths: AR models capture long-range dependencies and produce fluent, context-aware outputs, while diffusion models operate in continuous latent spaces to refine high-fidelity visual details. However, existing hybrids often lack systematic guidance on how and why to allocate model capacity between these paradigms. In this work, we introduce $\textit{MADFormer}$, a Mixed Autoregressive and Diffusion Transformer that serves as a testbed for analyzing AR-diffusion trade-offs. $\textit{MADFormer}$ partitions image generation into spatial blocks, using AR layers for one-pass global conditioning across blocks and diffusion layers for iterative local refinement within each block. Through controlled experiments on FFHQ-1024 and ImageNet, we identify two key insights: (1) block-wise partitioning significantly improves performance on high-resolution images, and (2) vertically mixing AR and diffusion layers yields better quality-efficiency balances---improving FID by up to 75\% under constrained inference compute. Our findings offer practical design principles for future hybrid generative models. Code and models will be released upon publication.

Large Language Models (LLMs) can compose poetry, but how far are they from human poets? In this paper, we introduce POEMetric, the first comprehensive framework for poetry evaluation, examining 1) basic instruction-following abilities in generating poems according to a certain form and theme, 2) advanced abilities of showing creativity, lexical diversity, and idiosyncrasy, evoking emotional resonance, and using imagery and literary devices, and 3) general appraisal of the overall poem quality and estimation of authorship. We curated a human poem dataset - 203 English poems of 7 fixed forms annotated with meter, rhyme patterns and themes - and experimented with 30 LLMs for poetry generation based on the same forms and themes of the human data, totaling 6,090 LLM poems. Based on POEMetric, we assessed the performance of both human poets and LLMs through rule-based evaluation and LLM-as-a-judge, whose results were validated by human experts. Results show that, though the top model achieved high form accuracy (4.26 out of 5.00, with Gemini-2.5-Pro as a judge; same below) and theme alignment (4.99), all models failed to reach the same level of advanced abilities as human poets, who achieved unparalleled creativity (4.02), idiosyncrasy (3.95), emotional resonance (4.06), and skillful use of imagery (4.49) and literary devices (4.67). Humans also defeated the best-performing LLM in overall poem quality (4.22 vs. 3.20). As such, poetry generation remains a formidable challenge for LLMs.

Computer-use agents (CUAs) hold the promise of performing a wide variety of general tasks, but current evaluations have primarily focused on simple scenarios. It therefore remains unclear whether such generalist agents can automate more sophisticated and specialized work such as software engineering (SWE). To investigate this, we introduce Programming with Pixels (PwP), the first comprehensive computer-use environment for software engineering, where agents visually control an IDE to perform diverse software engineering tasks. To enable holistic evaluation, we also introduce PwP-Bench, a benchmark of 15 existing and new software-engineering tasks spanning multiple modalities, programming languages, and skillsets. We perform an extensive evaluation of state-of-the-art open-weight and closed-weight CUAs and find that when interacting purely visually, they perform significantly worse than specialized coding agents. However, when the same CUAs are given direct access to just two APIs—file editing and bash operations—performance jumps, often reaching the levels of specialized agents despite having a task-agnostic design. Furthermore, when given access to additional IDE tools via text APIs, all models show further gains. Our analysis shows that current CUAs fall short mainly due to limited visual grounding and the inability to take full advantage of the rich environment, leaving clear room for future improvements. PwP establishes software engineering as a natural domain for benchmarking whether generalist computer-use agents can reach specialist-level performance on sophisticated tasks.

Large language models (LLMs) have achieved remarkable progress in code generation, yet their potential for software protection remains largely untapped. Reverse engineering continues to threaten software security, while traditional virtual machine protection (VMP) relies on rigid, rule-based transformations that are costly to design and vulnerable to automated analysis. In this work, we present the first protection-aware framework that learns robust representations of VMP-protected code. Our approach builds large-scale paired datasets of source code and normalized VM implementations, and introduces hierarchical dependency modeling at intra-, preceding-, and inter-instruction levels. We jointly optimize language modeling with functionality-aware and protection-aware contrastive objectives to capture both semantic equivalence and protection strength. To further assess resilience, we propose a protection effectiveness optimization task that quantifies and ranks different VM variants derived from the same source. Coupled with a two-stage continual pre-training and fine-tuning pipeline, our method enables models to generate, compare, and reason over protected code. Extensive experiments show that our framework significantly improves robustness across diverse protection levels, opening a new research direction for learning-based software defense. In this work, we present ShieldedCode, the first protection-aware framework that learns robust representations of VMP-protected code. Our method achieves 26.95% Pass@1 on L0 VM code generation compared to 22.58% for GPT-4o., and improves binary similarity detection Recall@1 by 10\% over state of art methods like jTrans.