Papers by Chao Zheng

31 papers
Graph Reasoning for Question Answering with Triplet Retrieval (2023.findings-acl)

Copied to clipboard

Challenge: Existing methods to answer complex questions require reasoning over knowledge graphs (KGs) state-of-the-art methods constrain retrieved knowledge in local subgraphs and discard more diverse triplets that are disconnected but useful for question answering.
Approach: They propose a method to retrieve the most relevant triplets from KGs and then rerank them, which are then concatenated with questions to be fed into language models.
Outcome: The proposed method outperforms state-of-the-art methods on commonsenseQA and OpenbookQA datasets with 4.6% absolute accuracy.
DNCASR: End-to-End Training for Speaker-Attributed ASR (2025.acl-long)

Copied to clipboard

Challenge: DNCASR is a novel end-to-end trainable system for speaker clustering and automatic speech recognition . it is designed to transcribe long multi-party meetings with speaker-attributed transcriptions .
Approach: They propose a joint neural speaker clustering and automatic speech recognition system . DNCASR uses two separate encoders to independently encode speaker characteristics and local waveform information .
Outcome: The proposed system outperforms a parallel system that does not have links between speaker and ASR decoders.
Hephaestus: Improving Fundamental Agent Capabilities of Large Language Models through Continual Pre-Training (2025.naacl-long)

Copied to clipboard

Challenge: Existing LLMs often rely on complex prompting or extensive fine-tuning to introduce new capabilities while preserving strong generalizability.
Approach: They propose a large-scale pre-training corpus to enhance LLM agents' capabilities . they use 103B agent-specific data encompassing 76,537 APIs .
Outcome: The proposed training corpus outperforms open-source LLMs and commercial LLM agents on three agent benchmarks.
Attention-Enhancing Backdoor Attacks Against BERT-based Models (2023.findings-emnlp)

Copied to clipboard

Challenge: Existing textual backdoor attacks focus on generating stealthy triggers or modifying model weights.
Approach: They propose a Trojan Attention Loss (TAL) which enhances the Trojan behavior by directly manipulating attention patterns.
Outcome: The proposed method improves the effectiveness of the backdoor attacks on different backbone models and tasks.
Automated Fine-Grained Mixture-of-Experts Quantization (2025.findings-acl)

Copied to clipboard

Challenge: specialized quantization framework for Mixture of Experts architectures is inadequate for model compression.
Approach: They propose a specialized quantization framework for Mixture of Experts architectures . they find that expert networks exhibit distinctive channel-wise outlier distributions ."
Outcome: The proposed framework improves on the Mixtral-8x7b-v0.1 architecture while maintaining minimal computational overhead.
Knowledge-Selective Pretraining for Attribute Value Extraction (2023.findings-emnlp)

Copied to clipboard

Challenge: Existing methods for AVE are limited on rare attributes due to poor generalization ability.
Approach: They propose to leverage pretraining and transfer learning to address weaknesses in existing methods.
Outcome: The proposed method achieves new state-of-the-art performance without pretraining on rare attributes with limited training resources.
Calibrated Speculative Decoding: Frequency-Guided Candidate Selection for Efficient Inference (2026.acl-long)

Copied to clipboard

Challenge: Speculative decoding (SD) is a powerful and efficient way to accelerate autoregressive generation.
Approach: They propose a training-free framework that recovers valid tokens discarded by standard verification . they use online correction memory and Semantic Consistency Gating to analyze rejections .
Outcome: The proposed framework outperforms existing methods and achieves peak throughput speedup of 2.33x.
Similarizing the Influence of Words with Contrastive Learning to Defend Word-level Adversarial Text Attack (2023.findings-acl)

Copied to clipboard

Challenge: Neural language models are vulnerable to word-level adversarial text attacks . previous word-based search methods assume important words influence prediction .
Approach: They propose a method for similarizing the influence of words with contrast learning that encourages model to learn sentence representations in which words of varying importance have a more uniform influence on prediction.
Outcome: The proposed method is compatible with various training methods and improves model robustness against various adversarial attacks.
DP3: Differentially Private Prompt Perturbation for Multi-turn LLM Inference (2026.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) are widely used for text understanding and generation . existing methods that assume single-turn interactions break down in multi-turn settings .
Approach: They propose a differentially private prompt perturbation framework for multi-turn LLM inference . DP3 constructs a perturbation mapping table to reuse perturbations for recurring tokens .
Outcome: The proposed framework reduces privacy costs and degrades cross-turn semantic coherence . it also provides a context-aware utility function to maintain semantic consistency across turns .
Traffic-R1: Reinforced LLMs Bring Human-Like Reasoning to Traffic Signal Control Systems (2026.acl-long)

Copied to clipboard

Challenge: Rapid urbanization and surging vehicle ownership intensify congestion . rapid urbanization drives crash rates, slow emergency response, and burden transit-poor communities .
Approach: They introduce a 3B-parameter foundation model with human-like reasoning for Traffic signal control (TSC) they use reinforcement learning and network communication to convert LLM into a traffic-control model that operates like a human traffic agent.
Outcome: The proposed model outperforms baselines and training-intensive RL controllers on a simulated traffic environment and reduces queue lengths by more than 5%.
Improving Consistency for Text Summarization with Energy Functions (2023.findings-emnlp)

Copied to clipboard

Challenge: Current abstractive summarization models generate inconsistent content due to the inherently noisy dataset and the discrepancy between maximum likelihood estimation based training objectives and consistency measurements.
Approach: They propose a new consistency taxonomy that categorizes inconsistent content into faithfulness, factuality, and self-supportiveness.
Outcome: Experiments on XSUM and CNN/DM datasets show that EnergySum mitigates the trade-off between accuracy and consistency.
Everyone is unique: Towards Behaviorally Heterogeneous Negotiation Dialogue Systems for Debt Collection (2026.acl-long)

Copied to clipboard

Challenge: Existing models that assume users to be static, rational agents with fixed preferences fail to capture rich behavioral heterogeneity in real-world debt collection scenarios.
Approach: They propose a public persona-enriched debt collection benchmark that highlights behavioral heterogeneity in negotiation.
Outcome: The proposed benchmark outperforms existing models in realistic scenarios using 16 state-of-the-art LLMs.
Geoparsing: Diagram Parsing for Plane and Solid Geometry with a Unified Formal Language (2026.findings-acl)

Copied to clipboard

Challenge: Recent advances in Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities across various vision reasoning tasks.
Approach: They propose a unified formal language that integrates plane and solid geometry, comprehensively covering geometric structures and semantic relations.
Outcome: The proposed language achieves state-of-the-art parsing performance and significantly boosts MLLMs’ capabilities for downstream geometry reasoning tasks.
Make Large Language Model a Better Ranker (2024.findings-emnlp)

Copied to clipboard

Challenge: Large Language Models (LLMs) demonstrate robust capabilities across various fields . current list-wise approaches fail in ranking tasks due to misalignment between ranking objectives and next-token prediction .
Approach: They propose a large language model framework with Aligned Listwise Ranking Objectives (ALRO) this framework provides explicit feedback in a listwise manner by introducing soft lambda loss .
Outcome: The proposed model outperforms existing recommendation methods and embedding-based recommendations without additional computational burdens.
TokenSelect: Efficient Long-Context Inference and Length Extrapolation for LLMs via Dynamic Token-Level KV Cache Selection (2025.emnlp-main)

Copied to clipboard

Challenge: Rapid advances in Large Language Models have spurred demand for processing extended context sequences . however, performance degradation due to sequence lengths out-of-distribution and excessively long inference times are limiting LLMs in long-context scenarios.
Approach: They propose a training-free method for efficient and accurate long-context inference . they selectively involves a few critical KV cache tokens in attention calculation .
Outcome: The proposed method speeds up attention computation and accelerates inference time while reducing selection overhead.
Learning from the Irrecoverable: Error-Localized Policy Optimization for Tool-Integrated LLM Reasoning (2026.acl-long)

Copied to clipboard

Challenge: Tool-integrated reasoning (TIR) enables LLM agents to solve tasks through planning, tool use, and iterative revision, but outcome-only reinforcement learning suffers from sparse, delayed rewards and weak step-level credit assignment.
Approach: They propose a tool-integrated reasoning approach that localizes the first irrecoverable step and leverages it for fine-grained credit assignment.
Outcome: The proposed algorithm outperforms strong Agentic RL benchmarks in math, science QA, and code execution with additional gains in Pass@K and Major@K scaling, rollout ranking quality, and tool-call efficiency.
Task-Agnostic Detector for Insertion-Based Backdoor Attacks (2024.findings-naacl)

Copied to clipboard

Challenge: Existing methods for textual backdoor detection are task-specific and less effective beyond sentence classification.
Approach: They propose a task-agnostic method for backdoor detection that leverages final layer logits and an efficient pooling technique.
Outcome: TABDet can jointly learn from diverse task-specific models, demonstrating superior detection efficacy over traditional methods.
Beyond Overlap Metrics: Rewarding Reasoning and Preferences for Faithful Multi-Role Dialogue Summarization (2026.findings-acl)

Copied to clipboard

Challenge: Existing methods for multi-role dialogue summarization favor surface-level imitation of references rather than genuine gains in faithfulness or alignment with human preferences.
Approach: They propose a framework that couples explicit cognitive-style reasoning with reward-based optimization for multi-role dialogue summarization.
Outcome: The proposed framework matches strong baselines on ROUGE and BERTScore, while in-depth analysis on SAMSum shows clear gains in factual faithfulness and model-based preference alignment.
Learning from the Dictionary: Heterogeneous Knowledge Guided Fine-tuning for Chinese Spell Checking (2022.findings-emnlp)

Copied to clipboard

Challenge: Chinese Spell Checking (CSC) aims to detect and correct Chinese spelling errors.
Approach: They propose a framework which renders Chinese Spell Checking model to learn heterogeneous knowledge from the dictionary in terms of phonetics, vision, and meaning.
Outcome: The proposed framework renders the CSC model to learn heterogeneous knowledge from the dictionary in terms of phonetics, vision, and meaning.
PARSE: An Efficient Search Method for Black-box Adversarial Text Attacks (2022.coling-1)

Copied to clipboard

Challenge: Neural networks are vulnerable to adversarial examples, i.e., under a black-box scenario.
Approach: They propose a word-level search algorithm that searches for subareas under dynamic search space following the subarea importance.
Outcome: The proposed algorithm can achieve comparable success rates to complex search methods while saving numerous queries and time.
Lost in the Context: Insufficient and Distracted Attention to Contexts in Preference Modeling (2025.acl-long)

Copied to clipboard

Challenge: Existing reward models concatenate contexts and responses, but they often ignore crucial segments of the context that are important for evaluating the response quality.
Approach: They propose a reward model that evaluates the response quality based on a given context and assigns a rewards reward.
Outcome: The proposed framework significantly improves preference modeling by increasing attention to relevant information within the context and achieves better generalizability.
SPEAK: Spiking Neurons as an Entropy-Aware Tokenizer for Large Language Models (2026.acl-long)

Copied to clipboard

Challenge: Existing tokenizers fail to explicitly leverage historical tokenization results . large language models (LLMs) have demonstrated remarkable effectiveness across NLP tasks .
Approach: They propose a tokenizer that integrates spiking neurons to explicitly leverage historical tokenization results.
Outcome: The proposed tokenizer leverages historical tokenization results, but does not selectively leverage history based on contextual relevance.
A Study of the Attention Abnormality in Trojaned BERTs (2022.naacl-main)

Copied to clipboard

Challenge: In computer vision, the trigger can be a fixed pattern overlaid on the images or videos.
Approach: They propose an attention-based Trojan detector to distinguish Trojaned models from clean ones by observing the attention focus drifting behavior of Trojanes.
Outcome: The proposed detector is based on transformer’s attention and can distinguish Trojan models from clean ones.
LongDocURL: a Comprehensive Multimodal Long Document Benchmark Integrating Understanding, Reasoning, and Locating (2025.acl-long)

Copied to clipboard

Challenge: Existing document understanding benchmarks only handle a small number of pages . existing models are limited to handling only a limited number of documents .
Approach: They propose a long document understanding benchmark that integrates three primary tasks and 20 sub-tasks based on different primary tasks.
Outcome: The proposed model outperforms existing benchmarks on open-source and closed-source models . the model outpersforms other models on more than 33,000 pages of documents .
Self-Adjust Softmax (2025.emnlp-main)

Copied to clipboard

Challenge: Usually, tokens with larger attention scores are important for the final prediction.
Approach: They propose to modify softmax(z) to z softmax and its normalized variant to improve the Transformer attention mechanism by making minor adjustments to the softmax function.
Outcome: The proposed model provides enhanced gradient properties compared to the vanilla softmax function.
Rethinking Word-level Adversarial Attack: The Trade-off between Efficiency, Effectiveness, and Imperceptibility (2024.lrec-main)

Copied to clipboard

Challenge: Neural language models have demonstrated impressive performance but remain vulnerable to word-level adversarial attacks.
Approach: They propose two standardized search spaces to address the problem of word-level adversarial attacks.
Outcome: The proposed search spaces improve performance and trade-offs in different scenarios.
Any Information Is Just Worth One Single Screenshot: Unifying Search With Visualized Information Retrieval (2025.acl-long)

Copied to clipboard

Challenge: Existing multimodal retrieval models are lacking in visual representations of multimodal data.
Approach: They propose a visualized information retrieval paradigm where multimodal information is represented by a unified visual format called Screenshots for various retrieval applications.
Outcome: The proposed model is based on a large dataset of screenshots from diverse sources . it is compared with existing models and lays a solid foundation for the new model .
scRAG: Hybrid Retrieval-Augmented Generation for LLM-based Cross-Tissue Single-Cell Annotation (2025.findings-acl)

Copied to clipboard

Challenge: Large language models (LLMs) have demonstrated impressive potential in a wide range of fields, including biology, genomics and healthcare.
Approach: They propose a framework that integrates advanced LLM-based RAG techniques into cross-tissue single-cell annotation.
Outcome: The proposed framework outperforms baseline models, generalist models, domain-specific methods, and trained classifiers on a cross-tissue dataset.
NSP-BERT: A Prompt-based Few-Shot Learner through an Original Pre-training Task —— Next Sentence Prediction (2022.coling-1)

Copied to clipboard

Challenge: Recent studies have shown that using prompts to utilize language models to perform downstream tasks is more effective than using token-level methods such as PET.
Approach: They propose to use a BERT original pre-training task abandoned by RoBERTa and other models to construct a sentence-level prompt-based method that does not need to fix the length of the prompt or the position to be predicted.
Outcome: The proposed method performs better than PET and EFL on a BERT pre-training task and is comparable to other prompt-based methods.
The Past Mistake is the Future Wisdom: Error-driven Contrastive Probability Optimization for Chinese Spell Checking (2022.findings-acl)

Copied to clipboard

Challenge: Chinese Spell Checking (CSC) aims to detect and correct spelling errors, which are caused by the phonological or visual similarity.
Approach: They propose an Error-driven COntrastive Probability Optimization framework to refine the knowledge representations of pre-trained language models to avoid predicting common characters.
Outcome: Extensive experiments and detailed analyses on SIGHAN datasets demonstrate that ECOPO is simple yet effective.

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