Papers by Yikun Zhang

14 papers
MARS2: Scaling Multi-Agent Tree Search via Reinforcement Learning for Code Generation (2026.acl-long)

Copied to clipboard

Challenge: Existing approaches to reinforcement learning are decoupled from structured search due to limited trajectory diversity.
Approach: They propose a unified RL framework that integrates multiple agents within a shared tree-structured search environment.
Outcome: Experiments show that MARS2 improves performance across diverse model combinations and training settings.
AgentCPM-GUI: Building Mobile-Use Agents with Reinforcement Fine-Tuning (2025.emnlp-demos)

Copied to clipboard

Challenge: Large language model agents have enabled GUI-based automation, but their deployment is limited by noisy data, poor generalization, and lack of support for non-English GUIs.
Approach: They propose an 8B-parameter GUI agent built for robust and efficient on-device GUI interaction.
Outcome: The proposed GUI agent achieves promising performance on five public benchmarks and proposed Chinese benchmark CAGUI.
Faithfully Explainable Recommendation via Neural Logic Reasoning (2021.naacl-main)

Copied to clipboard

Challenge: Existing models for explainable recommendation have neglected faithfulness of KG reasoning .
Approach: They propose to draw on interpretable logical rules to guide path-reasoning process for explanation generation.
Outcome: The proposed method delivers high-quality recommendations and ascertains the faithfulness of the derived explanation.
Uncertainty Aware Learning for Language Model Alignment (2024.acl-long)

Copied to clipboard

Challenge: Existing alignment strategies that focus on diverse and high-quality data often overlook the intrinsic uncertainty of tasks, learning all data samples equally.
Approach: They propose to introduce the sample uncertainty into the alignment of different task scenarios by a simple fashion by setting the label smoothing value of training according to the uncertainty of individual samples.
Outcome: The proposed model outperforms standard supervised fine-tuning on high-entropy tasks and complex low-entropic tasks.
VideoPro: Adaptive Program Reasoning for Long Video Understanding (2026.acl-long)

Copied to clipboard

Challenge: Existing methods for understanding long videos are limited due to the sparsity of visual evidence relevant to a given query.
Approach: They propose a framework that enables VideoLLMs to reason over long videos and refine their predictions through executable programs.
Outcome: The proposed framework outperforms existing methods across long-video understanding benchmarks.
Rescue: Ranking LLM Responses with Partial Ordering to Improve Response Generation (2024.acl-srw)

Copied to clipboard

Challenge: Customizing LLMs for a specific task involves separating high-quality responses from lower-quality ones. Obtaining a large volume of expert-annotated data is costly for most tasks.
Approach: They propose a method that trains the model to prioritize the best responses from a pool of candidates created for a task using ranking metrics.
Outcome: The proposed method is more robust, less sensitive to noise, and can be achieved with limited human annotations or through heuristic methods.
Think-Search-Patch: A Retrieval-Augmented Reasoning Framework for Repository-Level Code Repair (2025.emnlp-industry)

Copied to clipboard

Challenge: Large language models suffer from multiple-file coding scenarios with strong inter-file dependencies . experimental results show that large language models exhibit inadequate performance in multi-file scenarios .
Approach: They propose a retrieval-augmented reasoning framework for repository-level code repair . they use a dataset to generate standardized patches based on the key snippets .
Outcome: The proposed framework improves retrieval accuracy and repair success on SWE-bench Lite . it surpasses models with larger size in managing extensive code contexts and fixing bugs spanning across multiple files.
SeCuRepair: Semantics-Aligned, Curriculum-Driven, and Reasoning-Enhanced Vulnerability Repair Framework (2026.acl-long)

Copied to clipboard

Challenge: Existing methods for automating vulnerability repair suffer from syntactic overfitting . nvd published 49,230 Common Vulnerabilities and Exposures (CVE) records in 2025 alone .
Approach: They propose a semantic-aware reward framework that optimizes for code semantic equivalence rather than lexical mimicry.
Outcome: The proposed framework outperforms state-of-the-art frameworks on repository-level splits . it incorporates expert-aligned reasoning mechanism that grounds patch generation in structured diagnosis.
BLADE: Benchmarking Language Model Agents for Data-Driven Science (2024.findings-emnlp)

Copied to clipboard

Challenge: Language model-based agents can be used to conduct and support data-driven science, but evaluating them on open-ended tasks is challenging due to multiple valid approaches, partially correct steps, and different ways to express the same decisions.
Approach: They propose a benchmark to automatically evaluate agents’ multifaceted approaches to open-ended research questions.
Outcome: BLADE evaluates agents’ multifaceted approaches to open-ended research questions using data from 12 datasets and research questions drawn from existing scientific literature.
DecEx-RAG: Boosting Agentic Retrieval-Augmented Generation with Decision and Execution Optimization via Process Supervision (2025.emnlp-industry)

Copied to clipboard

Challenge: Recent advances in outcome-supervised reinforcement learning (RL) have shown strong performance, but this approach still suffers from inefficient exploration, sparse reward signals, and ambiguous global reward feedback.
Approach: They propose a model that models RAG as a Markov Decision Process (MDP) and introduces an efficient pruning strategy to optimize data expansion.
Outcome: The proposed model outperforms existing methods and achieves an average performance improvement of 6.2% across six datasets.
Beyond Self-Report: Bridging the Intention-Behavior Gap in Critical Thinking Assessment via Interpretable Multi-Agent System (2026.acl-long)

Copied to clipboard

Challenge: Accurate assessment of critical thinking is limited by the Intention Behavior Gap in psychology . evaluators that measure self-reported competence are limited by multiagent architectures .
Approach: They propose a framework that operationalizes cognitive assessment into an interpretable multi-agent workflow with Assessment Chain-of-Thought.
Outcome: The proposed framework aligns better with human expert ratings than gold-standard inventories on large-scale simulations and human participants.
Understanding the RoPE Extensions of Long-Context LLMs: An Attention Perspective (2025.coling-main)

Copied to clipboard

Challenge: Enabling LLMs to handle lengthy context is currently a research hotspot . a notable challenge limiting further customization is the inability of LLM to utilize context beyond pretrained length due to the inherent flaw of rotary position embedding (RoPE).
Approach: They propose to extend the RoPE from an attention perspective and on two benchmarking tasks.
Outcome: The proposed extension of the RoPE improves extrapolation and retrieval errors.
ZigZagKV: Dynamic KV Cache Compression for Long-context Modeling based on Layer Uncertainty (2025.coling-main)

Copied to clipboard

Challenge: Existing methods to accelerate inference of Large Language models (LLMs) are limited in their ability to retain key tokens as input length increases.
Approach: They propose a method that leverages layer uncertainty to allocate budget size for each layer to reduce memory usage.
Outcome: The proposed method reduces memory usage of the KV caches to only 20% when compared to full KV inference while achieving nearly lossless performance.
SecureVibeBench: Benchmarking Secure Vibe Coding of AI Agents via Reconstructing Vulnerability-Introducing Scenarios (2026.acl-long)

Copied to clipboard

Challenge: Existing benchmarks fail to capture scenarios in which vulnerabilities are introduced by humans . we evaluate 5 popular code agents supported by 5 LLMs on SecureVibeBench .
Approach: They propose a benchmarking tool that compares 105 C/C++ secure coding tasks . they use real-world open-source vulnerabilities and a comprehensive evaluation tool .
Outcome: The proposed benchmarks show that code agents struggle to produce correct and secure code . the best performing agent produces merely 23.8% correct and secured solutions .

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations