Papers by Yuxuan Sun

21 papers
Sparse-RL: Breaking the Memory Wall in LLM Reinforcement Learning via Stable Sparse Rollouts (2026.acl-long)

Copied to clipboard

Challenge: Existing methods for storing key-value caches during long-horizon rollouts cause performance collapses.
Approach: They propose a new training paradigm that empowers stable RL training under sparse rollouts.
Outcome: The proposed model reduces rollout overhead while maintaining the performance.
TritonBench: Benchmarking Large Language Model Capabilities for Generating Triton Operators (2025.findings-acl)

Copied to clipboard

Challenge: Triton is a high-level Python-like programming language for building efficient GPU kernels.
Approach: They propose a TritonBench benchmark that provides a comprehensive evaluation of Tritonic operators on widely deployed GPUs.
Outcome: The proposed benchmarks show that current LLMs struggle to generate efficient Triton operators on widely deployed GPUs aligned with industry applications.
On LLM-Based Scientific Inductive Reasoning Beyond Equations (2025.emnlp-main)

Copied to clipboard

Challenge: Existing research on inductive reasoning models emphasizes rule design without grounding them in specific scenarios.
Approach: They propose to use LLMs to learn underlying patterns from limited examples in entirely new environments.
Outcome: The proposed benchmark evaluates the inductive reasoning abilities of large language models in scientific settings.
HD-Eval: Aligning Large Language Model Evaluators Through Hierarchical Criteria Decomposition (2024.acl-long)

Copied to clipboard

Challenge: Large language models (LLMs) are a promising alternative to expensive human evaluations.
Approach: They propose a framework that iteratively aligns LLM-based evaluators with human preference . they decompose a given evaluation task into finer-grained criteria .
Outcome: The proposed framework iteratively aligns LLM-based evaluators with human preference . it decomposes a given evaluation task into finer-grained criteria . the framework is efficient to train and more explainable than relying solely on prompts .
PersLLM: A Personified Training Approach for Large Language Models (2025.findings-emnlp)

Copied to clipboard

Challenge: Large language models exhibit human-like intelligence, enabling them to simulate human behavior and support various applications that require both humanized communication and extensive knowledge reserves.
Approach: They propose a framework for better data construction and model tuning to unlock the potential of LLM personification by using Chain-of-Thought prompting and anti-induction.
Outcome: The proposed framework improves data construction and model tuning for insufficient data usage and rigid behavior patterns.
StorySparkQA: Expert-Annotated QA Pairs with Real-World Knowledge for Children’s Story-Based Learning (2024.emnlp-main)

Copied to clipboard

Challenge: Existing story reading systems fail to capture the nuances of how education experts think when conducting interactive story reading activities.
Approach: They propose to use existing question-answering (QA) datasets to capture experts' annotations and thinking process to construct a story-based annotation framework.
Outcome: The proposed framework captures experts’ annotations and thinking process and can be used to generate 5, 868 expert-annotated QA pairs with real-world knowledge.
KBAlign: Efficient Self Adaptation on Specific Textual Knowledge Bases (2025.findings-emnlp)

Copied to clipboard

Challenge: Existing methods for retrieval-augmented generation (RAG) are limited and fine-tuning incurs prohibitive costs of external signals.
Approach: They propose a self-supervised framework that enhances RAG systems through efficient model adaptation.
Outcome: The proposed framework achieves 90% of the performance gain obtained through GPT-4-supervised adaptation while relying entirely on self-annotation of much smaller models.
Cross-Lingual Knowledge Editing in Large Language Models (2024.acl-long)

Copied to clipboard

Challenge: Knowledge editing is a promising technique to adapt large language models to new knowledge without retraining from scratch.
Approach: They propose to use a multilingual dataset to translate a large-scale cross-lingual synthetic dataset from English to Chinese and then to evaluate their performance in Chinese.
Outcome: The proposed method can change model performance on several special cases without retraining from scratch.
Democratizing Reasoning Ability: Tailored Learning from Large Language Model (2023.emnlp-main)

Copied to clipboard

Challenge: Large language models (LLMs) exhibit impressive emergent abilities in natural language processing, but their democratization is hindered due to huge computation requirements and closed-source nature.
Approach: They propose a tailored learning approach to distill the exclusive reasoning ability to smaller LMs to facilitate democratization.
Outcome: The proposed approach enables the democratization of the exclusive reasoning ability by leveraging the black-box model as a reasoning teacher.
TestAgent: An Adaptive and Intelligent Expert for Human Assessment (2025.findings-acl)

Copied to clipboard

Challenge: Existing adaptive testing methods face several challenges due to mechanized nature of most algorithms and noisy response data.
Approach: They propose to use large language models to enhance adaptive testing through interactive engagement to capture test-takers’ responses and anomalies.
Outcome: The proposed agent achieves more accurate results with 20% fewer questions than state-of-the-art baselines and testers preferred it in speed, smoothness, and other dimensions.
FR-Spec: Accelerating Large-Vocabulary Language Models via Frequency-Ranked Speculative Sampling (2025.acl-long)

Copied to clipboard

Challenge: Speculative sampling is an efficient way to accelerate the auto-regressive generation process of large language models.
Approach: They propose a frequency-ranked speculative sampling framework that optimizes draft candidate selection through vocabulary space compression.
Outcome: Experiments show that FR-Spec reduces LM Head computation overhead by 75% while ensuring the equivalence of the final output distribution.
AutoReproduce: Automatic AI Experiment Reproduction with Paper Lineage (2026.acl-long)

Copied to clipboard

Challenge: Efficient reproduction of research papers requires deep domain expertise.
Approach: They propose a framework that systematically mines implicit knowledge from the cited literature to reproduce experimental code in a complete, end-to-end manner.
Outcome: The proposed framework surpasses baselines across all metrics and reproduces experimental code in a complete, end-to-end manner.
MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmark (2025.acl-long)

Copied to clipboard

Challenge: Recent advances in multimodal large language models have led to progress in tackling complex reasoning tasks that combine textual and visual information.
Approach: They introduce a robust version of the Massive Multi-discipline Multimodal Understanding and Reasoning (MMMU) benchmark.
Outcome: The proposed model performs lower on MMMU-Pro than on the previous benchmark, ranging from 16.8% to 26.9%.
OPeRA: A Dataset of Observation, Persona, Rationale, and Action for Evaluating LLMs on Human Online Shopping Behavior Simulation (2026.acl-long)

Copied to clipboard

Challenge: evaluating LLMs' ability to mimic real user behavior remains an open challenge due to the lack of high-quality, publicly available datasets that capture both the observable actions and the internal reasoning of an actual user.
Approach: They propose a dataset of Observation, Persona, Rationale, and Action collected from real human participants during online shopping sessions.
Outcome: The proposed dataset is the first to evaluate how well current LLMs can accurately simulate the next web action of a specific user.
An Empirical Study of Many-to-Many Summarization with Large Language Models (2025.acl-long)

Copied to clipboard

Challenge: Recent studies have shown that large language models (LLMs) have strong multilingual abilities, giving them the potential to perform M2MS in real applications.
Approach: They propose to use many-to-many summarization (M2MS) to generate a brief summary in any language given a document also in any other language.
Outcome: The proposed model outperforms zero-shot LLMs in terms of automatic evaluations.
RIVA: A Pre-trained Tweet Multimodal Model Based on Text-image Relation for Multimodal NER (2020.coling-main)

Copied to clipboard

Challenge: Named entity recognition (MNER) for tweets is a key task of many applications.
Approach: They propose a pre-trained multimodal named entity recognition model based on Relationship Inference and Visual Attention (RIVA) for tweets.
Outcome: The proposed model improves on the multimodal named entity recognition (MNER) task on tweets with the aid of visual clues.
SWE-Mutation: Can LLMs Generate Reliable Test Suites in Software Engineering? (2026.findings-acl)

Copied to clipboard

Challenge: Evaluating software engineering capabilities is a core component of large language models (LLMs).
Approach: They propose a benchmark to evaluate LLM-generated test suites that introduces mutated solutions that attempt to "fool" them.
Outcome: The proposed test suites are based on 2,636 mutated variants derived from 800 original instances and include a multilingual subset spanning nine programming languages.
Optimizing Native Sparse Attention with Latent Attention and Local Global Alternating Strategies (2026.findings-acl)

Copied to clipboard

Challenge: Existing research has proposed a variety of training-free and post-training methods for selecting critical key-value pairs at each generation step.
Approach: They propose to use local (sliding-window) and global (compression/selective) attention across layers to enlarge long-context modeling.
Outcome: Experiments on models from 340M to 1.3B parameters show that the proposed method matches or exceeds full attention and native sparse attention in both common-sense reasoning and long-context understanding tasks.
AutoVecCoder: Teaching LLMs to Generate Explicitly Vectorized Code (2026.findings-acl)

Copied to clipboard

Challenge: Current development practices face a dichotomy between automation and performance.
Approach: They propose a framework to empower LLMs with the capability of automated explicit vectorization.
Outcome: The proposed framework achieves state-of-the-art performance on the SSE and AVX subsets of SimdBench.
Calibrating LLM-Based Evaluator (2024.lrec-main)

Copied to clipboard

Challenge: Existing models for large language models lack the ability to calibrate their outputs towards human preference.
Approach: They propose a multi-stage, gradient-free approach to calibrate an LLM-based evaluator toward human preference.
Outcome: The proposed approach improves correlation with expert evaluation on multiple text quality evaluation datasets.

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