arXiv subjects

Zhaochun Ren

Publications and source records attributed to Zhaochun Ren.

At least 19 recordsLinked to original sources

arXiv2026

EviACT: An Evidence-to-Action Framework for Agentic Program Repair

LLM-based agents have moved automated program repair (APR) from fixed-context patch generation to interactive repository-level repair. However, existing agentic APR systems still struggle to use execution evidence to guide localization, patch generation, and validation. We propose EviACT (Evidence-to-Action), an agentic APR framework that coordinates three evidence-driven guardrails across repair stages. The retrieval scaffold grounds repair context, the compile gate filters invalid edits, and the test-driven gate checks target-test recovery before full regression. Across four benchmarks, EviACT improves resolve rate over the strongest reported comparable baselines by 1.6-6.0 percentage points and shows 70.1-88.6% lower reported per-bug API cost where baseline costs are available. Ablations and diagnostics suggest that these gains are associated with the coordinated evidence-to-action chain, making agentic APR more effective and efficient.

cs.SE↗

arXiv2026

SearchJev: A Calibrated System-1 Model for Search Agents

Search agents repeatedly make short decisions about relevance, evidence sufficiency, and search actions. Using generative language models for these decisions introduces latency and unreliable confidence. We present SearchJev, a calibrated System-1 model that separates search decisions from System-2 reasoning and generation. Given a search state and a decision schema, SearchJev scores legal options without autoregressive output generation. We propose Soft-Label Learning for Calibrated Decisions (SLCD) to learn decision probabilities from uncertain supervision and improve the reliability of decision confidence. In a dual-system search agent, SearchJev handles short decisions and delegates uncertain judgments to System 2, which retains planning, query generation, and answer composition. We also introduce SearchDecision-Bench, a benchmark unifying six types of search decisions for training and evaluation. On SearchDecision-Bench, SearchJev improves decision quality over same-size Qwen3.5 autoregressive models, achieves 5.2-5.3 times faster decisions, and reduces average expected calibration error by 41-74%. On BrowseComp-Plus, the dual-system agents achieve a 3.7-4.7 times speedup in active search time while improving answer accuracy from 45% to up to 54%.

cs.IR↗

arXiv2026

UserArena: Benchmarking Interactive User Simulation for Long-Horizon Shopping Agents

Agentic recommendation pursues shopper goals through sequences of context-dependent decisions. Developing and evaluating such policies requires faithful recommendation world models that capture both the interactive shopping environment and users' behavioral responses over long interactions. Existing resources rarely provide long-horizon, fine-grained real-user trajectories or support trajectory-level evaluation and offline validation against production A/B results. We introduce UserArena, a real-world e-commerce interactive benchmark containing 64,729 session trajectories from 11,010 users, spanning 566,102 items and 13 action types. UserArena pairs these trajectories with an interactive environment and supports offline validation against outcomes from a controlled production A/B test, enabling five-level evaluation from output validity and local decisions to sequential behavior, personalization, and outcomes. To improve simulator fidelity over these long-horizon interactions, we further propose On-Policy Trajectory Distillation (OPTD), which uses trajectory-level intent to provide dense teacher targets on student-generated rollouts. Experiments show that OPTD achieves the strongest overall fidelity among the evaluated simulators under UserArena's evaluation protocol and most closely reproduces the production policy effect in offline simulation.

cs.IR↗

arXiv2026

SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm

Recursive self-improvement (RSI) in scientific research aims to improve agents' research capabilities through their own experimental experience. Scientific agent swarms can combine evidence and implementations from different research directions to guide their agents' subsequent experiments. However, poor experimental results can leave unclear whether to revise a hypothesis or its implementation, while overall program scores can hide components useful to other agents. To address these challenges, we introduce SwarmRSI, a framework that evolves scientific agent swarms for recursive self-improvement. We devise Feedback-Guided Swarm Mutation to refine hypotheses and their implementations using experimental feedback from other agents. We also propose Complementary Swarm Recombination to reuse retained hypotheses and complementary program components across agents, allowing partial results to support new research. Recursive Swarm Evolution integrates both mechanisms into a feedback loop, using new experimental evidence to guide subsequent research. Experiments on OpenProblems, NatureBench and MLE-Bench Lite demonstrate SwarmRSI's state-of-the-art (SOTA) performance. Further evaluations show that accumulated research experience improves performance on held-out tasks. Biologists also reproduce its programs for follow-up biological analyses.

cs.AI↗

arXiv2026

Measuring Concept Content in Text from LLM Activations: ESG Evidence from Concept Vectors and Linear Probes

Recent work shows that Large Language Models (LLMs) encode more information in their activations than they show in their responses. We ask whether the information contained in activations could be used as a source of measurement for the presence of concepts. We compare two algorithms for extraction of concepts, i.e., Recursive Feature Machine (RFM) algorithm and linear probing against an embedding baseline, surface baselines, and the model's own response using a human-annotated Environmental, Social and Governance (ESG) dataset of financial texts. The best linear probe ranges from 0.6 to 3.3 percentage points below fine-tuned domain classifiers. The probes also outscore the model's own response in all comparisons. Furthermore, the probes outperform the RFM concept vectors in all accuracy comparisons. In a synthetic mixture test, their mean-pooled score tracks the number of concept sentences in a paragraph better than the concept-vector score. Put together, a linear probe on a frozen LLM is a cost-effective and generalizable alternative to fine-tuning a domain classifier.

cs.CL↗

arXiv2026

DeltaLLM: Delta-Key Sparse Attention for Edge LLMs

On-device LLM inference is increasingly important for edge platforms but remains challenging due to high computational and memory demands and tight resource constraints. Modern applications often involve long prompts. Although compact models reduce linear-layer costs through weight compression, attention remains largely unoptimized and becomes the dominant bottleneck, especially during prefilling. Existing attention pruning methods primarily target highly parallel hardware (e.g., GPUs) and long context windows, limiting their suitability for edge scenarios. We present DeltaLLM, a training-free LLM framework based on algorithm-hardware-dataflow co-design to accelerate prefilling attention. DeltaLLM introduces a causal-attention-aware K-only delta formulation that enables Delta-Key reuse across query heads under Grouped-Query Attention (GQA). A context-aware hybrid strategy preserves exact attention near the diagonal while applying delta approximation to nonlocal regions. For score-value multiplication, we propose a vectorized structured sparse online softmax (VSSOS) that converts fine-grained spatial score sparsity into hardware-friendly column-level skipping. A specialized vector-processor architecture and unified sparse/dense attention dataflow exploit the resulting sparsity. Evaluated on BitNet-b1.58-2B-4T and LLaMA3.2-1B-Instruct, DeltaLLM achieves up to 60% Delta-Key sparsity and 63% structured attention-score sparsity with negligible average accuracy loss. On the KV260 FPGA board, it delivers a 1.62x speedup over TeLLMe-v2 while reducing DSP and BRAM usage. These results show that DeltaLLM supports efficient edge LLM deployment and integration with existing pipelines without fine-tuning.

cs.AI↗

arXiv2026

EvoScientist: Towards Multi-Agent Evolving AI Scientists for End-to-End Scientific Discovery

The increasing adoption of Large Language Models (LLMs) has enabled AI scientists to perform complex end-to-end scientific discovery tasks requiring coordination of specialized roles, including idea generation and experimental execution. However, most state-of-the-art AI scientist systems rely on static, hand-designed pipelines and fail to adapt based on accumulated interaction histories. As a result, these systems overlook promising research directions, repeat failed experiments, and pursue infeasible ideas. To address this, we introduce EvoScientist, an evolving multi-agent AI scientist framework that continuously improves research strategies through persistent memory and self-evolution. EvoScientist comprises three specialized agents: a Researcher Agent (RA) for scientific idea generation, an Engineer Agent (EA) for experiment implementation and execution, and an Evolution Manager Agent (EMA) that distills insights from prior interactions into reusable knowledge. EvoScientist contains two persistent memory modules: (i) an ideation memory, which summarizes feasible research directions from top-ranked ideas while recording previously unsuccessful directions; and (ii) an experimentation memory, which captures effective data processing and model training strategies derived from code search trajectories and best-performing implementations. These modules enable the RA and EA to retrieve relevant prior strategies, improving idea quality and code execution success rates over time. Experiments show that EvoScientist outperforms 7 open-source and commercial state-of-the-art systems in scientific idea generation, achieving higher novelty, feasibility, relevance, and clarity via automatic and human evaluation. EvoScientist also substantially improves code execution success rates through multi-agent evolution, demonstrating persistent memory's effectiveness for end-to-end scientific discovery.

cs.CL↗

arXiv2026

Beyond Refusal Patterns: Safe-Role Internalization for Robust and Generalizable LLM Safety Alignment

Large Language Models (LLMs) have achieved remarkable capabilities but remain vulnerable to jailbreak attacks that elicit harmful or unsafe outputs. Existing safety alignment approaches, including Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF), often require substantial attack-specific supervision and computational resources, while remaining susceptible to shallow safety alignment and over-refusal. To address these challenges, we introduce SSRFT(Supervised Safe-Role Fine-Tuning), the first framework that reformulates safety alignment as the internalization of a predefined safe role. SSRFT constructs a Safe-Role Question-Answer (SRQA) dataset from psychometric questions, limited jailbreak prompts, and a safe-role description. Role-consistent responses are synthesized, validated, and expanded into diverse scenarios, enabling models to internalize safety-oriented values and principles rather than explicit refusal patterns. Experiments across multiple Base and Instruct models show that SSRFT achieves more robust and generalizable safety alignment than standard SFT. SSRFT shows substantially greater robustness to prefilling attacks and better generalization to unseen jailbreak domains, while reducing over-refusal on benign queries and preserving the model's general capabilities. These results establish safe-role internalization as an effective alternative to refusal-centric safety alignment. Warning: This paper contains examples of harmful and toxic language.

cs.AI↗

arXiv2026

Optimizing VLP-aligned Multimodal Intent Representation with Correct Visual Instantiation for Zero-Shot Composed Image Retrieval

ZS-CIR aims to retrieve a target image from a reference image and a modification text without paired supervision, typically by encoding composed queries as text-dominant representations within the image-text matching space of VLPs. However, queries reconstructed by visual pseudo-word learning or MLLM-based target reasoning often deviate from the native VLP representation space due to reference noise and coarse text fusion in the former, and verbose, weakly visually grounded descriptions in the latter. In this paper, we propose a unified ZS-CIR framework (named VMIR-CVI) to reconstruct multimodal composite queries from two complementary perspectives for optimizing VLP-compatible multimodal intent representation. First, it reasons and converts the multimodal intent into a unified textual description, aligning with the native text space of the VLP backbones to produce more retrieval-compatible textual queries. Second, it reconstructs the query representation with correctly decoupled visual instance cues, reducing reference noise while preserving target-relevant content. Specifically, a VLP-aligned Multimodal Intent Reasoning (VMIR) module injects few-shot VLP-style exemplars into chain-of-thought prompts, guiding the MLLM to generate target-consistent intent queries. A Training-free Visual Instance Disentanglement (TVID) module decouples fine-grained visual instances from global reference features without additional optimization. Finally, a lightweight Hybrid-modal Intent Alignment and Fusion (HIAF) module integrates the reasoned textual intent and disentangled visual cues into a unified hybrid-modal representation for robust ZS-CIR. Extensive experiments on three CIR benchmarks, namely CIRR, CIRCO and FashionIQ, show that VMIR-CVI significantly outperforms existing baselines and achieves new state-of-the-art performance. Code and trained models will be publicly released.

cs.IR↗

arXiv2026

Beyond Imitation: Reflective On-Policy Self-Distillation for LLM Reasoning

On-policy self-distillation (OPSD) improves the reasoning capabilities of large language models (LLMs) by providing dense token-level supervision for on-policy rollouts. However, existing OPSD methods often yield limited gains on complex reasoning tasks and suffer from severe training instability. We identify two key causes: conditioning the self-teacher on a complete verified solution encourages imitation of complete reference trajectories rather than extraction of transferable reasoning insights, while indiscriminate full-response distillation imposes superfluous supervision on already-valid reasoning prefixes. Together, these issues suppress reasoning diversity and contribute to late-stage mode collapse. We propose Reflective On-policy Self-Distillation (ROSD), which distills transferable reasoning insights rather than complete reference trajectories. For each erroneous rollout, a self-reflector contrasts it with a correct rollout from the same group to derive a corrective idea and identify the sentence containing the first reasoning error. The corrective idea provides the self-teacher with targeted guidance, while the diagnosed error boundary allows ROSD to mask out the distillation loss over the valid prefix and apply token-level distillation only from the first erroneous sentence onward. Experiments across multiple reasoning benchmarks and model backbones show that ROSD consistently outperforms standard OPSD and reinforcement learning baselines, better preserves reasoning diversity, stabilizes training, and mitigates late-stage mode collapse. Code is available at https://github.com/ZiqiZhao1/ROSD.

cs.CL↗

arXiv2026

Compression Beyond the Uncompressed: A Two-Stage Training Recipe for Soft Context Compression in RAG

Retrieval-Augmented Generation (RAG) improves knowledge-intensive generation by conditioning language models on retrieved documents, but processing these documents becomes increasingly expensive as retrieval depth grows. Soft context compression reduces this cost by encoding documents into compact continuous representations that can be precomputed and reused across queries. However, many existing methods train compressed models by distilling from a full-context teacher. When the teacher is wrong, such distillation can reinforce its errors, while teacher imitation provides no direct signal for improving beyond the teacher. We propose DEX-Comp, a two-stage training recipe that separates reliable imitation from targeted exploration. Pure Distillation learns only from teacher-correct questions to mitigate error propagation, while Hard Exploration applies outcome-based reinforcement learning to teacher-failed questions to directly optimize answer correctness. Across five open-domain QA benchmarks and retrieval depths from top-$5$ to top-$30$, DEX-Comp at $16\times$ compression outperforms all evaluated compression baselines and surpasses the untuned full-context RAG model in average accuracy, while reducing time-to-first-token by $4.4\times$--$23.7\times$. Evaluations across additional datasets and backbones further demonstrate its generalization.

cs.CL↗

arXiv2026

What Makes a Good Semantic ID for Generative Recommendation? A Reproducibility Study

Generative recommendation has emerged as an active research direction, where items are commonly represented by semantic IDs (SIDs): discrete codes generated token by token. Despite strong empirical results, SID designs vary widely in construction strategy, codebook organization, and code length, making their true impact on recommendation performance unclear. We conduct a large-scale reproducibility study to systematically investigate the impact of semantic ID design on generative recommendation under a unified experimental framework. We focus on a fundamental question: What makes a good semantic ID for generative recommendation? To answer this question, we examine four aspects: the relative effectiveness of different semantic ID designs, the connection between codebook utilization and recommendation quality, the effect of semantic code length, and the influence of semantic ID design on local item semantic preservation. Through a unified evaluation and additional cross-dataset controlled analyses, we find that the effects of SID design are largely non-monotonic: no single SID design is universally best, and commonly used RQ-VAE- and OPQ-based designs can behave inconsistently across datasets. The method with the most balanced first-level codebook is not consistently the best recommender, showing that utilization is diagnostic but insufficient. Scaling either the generative backbone or the SID length is also not always beneficial. Finally, semantic-neighborhood analysis reveals that no single SID design dominates all notions of local semantic preservation; instead, different designs exhibit complementary strengths that remain stable across datasets and neighborhood sizes. Our study provides a controlled and reproducible understanding of semantic ID design and offers practical insights for future generative recommender systems.

cs.IR↗

arXiv2026

A Token is Worth over 1,000 Tokens: Efficient Knowledge Distillation through Low-Rank Clone

Training high-performing Small Language Models (SLMs) remains costly, even with knowledge distillation and pruning from larger teacher models. Existing work often faces three key challenges: (1) information loss from hard pruning, (2) inefficient alignment of representations, and (3) underutilization of informative activations, particularly from Feed-Forward Networks (FFNs). To address these challenges, we introduce Low-Rank Clone (LRC), an efficient pre-training method that constructs SLMs aspiring to behavioral equivalence with strong teacher models. LRC trains a set of low-rank projection matrices that jointly enable soft pruning by compressing teacher weights, and activation clone by aligning student activations, including FFN signals, with those of the teacher. This unified design maximizes knowledge transfer while removing the need for explicit alignment modules. Extensive experiments with open-source teachers (e.g., Llama-3.2-3B-Instruct, Qwen2.5-3B/7B-Instruct) show that LRC matches or surpasses state-of-the-art models trained on trillions of tokens--while using only 20B tokens, achieving over 1,000x training efficiency. Our codes and model checkpoints are available at https://github.com/CURRENTF/LowRankClone and https://huggingface.co/collections/JitaiHao/low-rank-clone-lrc-6828389e96a93f1d4219dfaf.

cs.CL↗

arXiv2026

Token-Level Credit Assignment Optimization for Generative Document Retrieval

Generative retrieval models perform document retrieval by autoregressively generating document identifiers (DocIDs). This process naturally forms a sequential decision problem, i.e., the model makes a sequence of token-level decisions, selecting a DocID token at each decoding step, with the resulting complete sequence identifying the retrieved document. However, relevance feedback is available only after the complete DocID has been generated and mapped to a document, resulting in a granularity mismatch between token-level generation decisions and document-level retrieval supervision. Consequently, existing reinforcement learning methods for generative retrieval rely on sequence-level rewards, assigning the same document-level relevance signal to every decoding step. Such uniform credit assignment obscures the contribution of individual token decisions, making it difficult to identify which decisions contribute to retrieval success or failure. In this paper, we propose Token-Level Credit Assignment for Generative Retrieval (TCA), a fine-grained reinforcement learning framework that aligns the granularity of credit assignment with that of autoregressive DocID generation. Unlike assigning a single reward to an entire generated DocID, TCA derives fine-grained rewards by comparing the hidden-state trajectory of each generated DocID with the gold DocID trajectory obtained from a frozen reference model. These trajectory-based rewards provide differentiated feedback across decoding steps, allowing the policy to reinforce generation paths that remain aligned with the target DocID. Moreover, TCA decouples token-level credit assignment from policy optimization and can be instantiated with both GRPO and PPO. Experiments on benchmarks show that our method consistently outperforms baselines, demonstrating the effectiveness of fine-grained supervision for aligning DocID generation.

cs.IR↗

arXiv2026

Closing the Indexing-Decoding Gap in Multimodal Generative Retrieval via Prefix Retention Optimization

Multimodal generative retrieval formulates multimodal retrieval as discrete identifier generation, eliminating the need for explicit similarity search over external embeddings. Existing approaches construct identifiers via residual quantization and decode them with trie-constrained beam search. This combination introduces an indexing-decoding gap: identifier learning objectives, including reconstruction and contrastive losses, do not explicitly enforce prefix discriminability during decoding. As a result, even well-optimized identifiers can be irreversibly pruned early in beam search due to low-rank prefixes. We theoretically characterize this gap and derive a survival bound that relates prefix retention to three controllable factors in indexing and decoding. Building on this bound, we propose PRO, prefix retention optimization, a unified framework comprising three mechanisms: (i) prefix ranking distillation aligns quantized prefix rankings with those induced by pre-quantization embeddings using a listwise loss; (ii) vocabulary scheduling increases codebook sizes from shallow to deep residual quantization levels to reduce early competition from non-target prefixes; and (iii) geometric score fusion vectorizes each candidate prefix and incorporates its similarity to the query into beam search scoring, further reducing the indexing-decoding mismatch. Experiments on nine multimodal retrieval tasks show that PRO improves retention of target identifier prefixes and outperforms existing multimodal generative retrieval baselines.

cs.IR↗

arXiv2026

LEGO Co-builder: Exploring Fine-Grained Vision-Language Modeling for Multimodal LEGO Assembly Assistants

Vision-language models (VLMs) are facing the challenges of understanding and following multimodal assembly instructions, particularly when fine-grained spatial reasoning and precise object state detection are required. In this work, we explore LEGO Co-builder, a hybrid benchmark combining real-world LEGO assembly logic with programmatically generated multimodal scenes. The dataset captures stepwise visual states and procedural instructions, allowing controlled evaluation of instruction-following, object detection, and state detection. We introduce a unified framework and assess leading VLMs such as GPT-4o, Gemini, and Qwen-VL, under zero-shot and fine-tuned settings. We also evaluated the framework using a reasoning-focused model, GLM-4.1-thinking. Our results show that while object detection achieved high performance (98.16% with fine-tuned InstructBLIP), fine-grained scene understanding and assembly state detection remain challenging: Fine-tuned MiniGPT-v2 reached only 37.52% F1 for identifying theme entities, and even advanced models such as GPT-4o achieved just 40.54% F1 on state detection. This highlights gaps in fine-grained visual understanding among existing models. We release the benchmark, codebase, and generation pipeline to support future research on multimodal assembly assistants grounded in real-world workflows.

cs.AI↗

arXiv2026

OpenReward: Learning to Reward Long-form Agentic Tasks via Reinforcement Learning

Reward models (RMs) have become essential for aligning large language models (LLMs), serving as scalable proxies for human evaluation in both training and inference. However, existing RMs struggle on knowledge-intensive and long-form tasks, where evaluating correctness requires grounding beyond the model's internal knowledge. This limitation hinders them from reliably discriminating subtle quality differences, especially when external evidence is necessary. To address this, we introduce OpenRM, a tool-augmented long-form reward model that systematically judges open-ended responses by invoking external tools to gather relevant evidence. We train OpenRM with Group Relative Policy Optimization (GRPO) on over 27K synthesized pairwise examples generated through a controllable data synthesis framework. The training objective jointly supervises intermediate tool usage and final outcome accuracy, incentivizing our reward model to learn effective evidence-based judgment strategies. Extensive experiments on three newly-collected datasets and two widely-used benchmarks demonstrate that OpenRM substantially outperforms existing reward modeling approaches. As a further step, we integrate OpenRM into both inference-time response selection and training-time data selection. This yields consistent gains in downstream LLM alignment tasks, highlighting the potential of tool-augmented reward models for scaling reliable long-form evaluation.

cs.CL↗

arXiv2026

ReportLogic: Evaluating Logical Quality in Deep Research Reports

Users increasingly rely on Large Language Models (LLMs) for Deep Research, using them to synthesize diverse sources into structured reports that support understanding and action. In this context, the practical reliability of such reports hinges on logical quality: whether the report's claims and arguments are explicitly supported and can be trusted as a basis for downstream use, rather than merely appearing fluent or informative. However, current evaluation frameworks largely overlook this requirement. To bridge this gap, we introduce ReportLogic, a benchmark that quantifies report-level logical quality through a reader-centric lens of auditability. Specifically, ReportLogic adopts a hierarchical taxonomy that evaluates whether readers can (1) trace an on-topic report structure with a unified analytical arc (Macro-Logic), (2) understand the progression with necessary context (Expositional-Logic), and (3) verify conclusions via explicit claim--support (Structural-Logic). Based on this taxonomy, we construct a human-annotated rubric-guided dataset and train an open-source LogicJudge for scalable evaluation. We further evaluate judge robustness via adversarial attacks, showing that off-the-shelf LLM judges are frequently influenced by superficial cues (e.g., verbosity), and reasoning modes can mask broken support relations. Overall, our results provide actionable guidance for building more robust logic evaluators and improving the logical reliability of LLM-generated reports.

cs.CL↗