Papers by Hanqi Yan
CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation (2025.emnlp-main)
Copied to clipboard
| Challenge: | Prior implicit CoT methods have underperformed in terms of efficiency and robustness by relying on natural language tokens for reasoning. |
| Approach: | They propose a training framework that compresses natural language CoT into continuous space by aligning hidden states of a designated token. |
| Outcome: | The proposed framework outperforms the existing state-of-the-art in 3.1x compression rate and 28.2% accuracy on GSM8k scale. |
The Mystery of In-Context Learning: A Comprehensive Survey on Interpretation and Analysis (2024.emnlp-main)
Copied to clipboard
| Challenge: | In-context learning (ICL) is a capability that enables large language models to excel in proficiency through demonstration examples. |
| Approach: | They present a survey on the interpretation and analysis of in-context learning . they focus on theoretical and empirical perspectives on the concept . |
| Outcome: | The proposed model can perform tasks with minimal examples without re-training and has demonstrated proficiency across various tasks with a minimal set of task-oriented examples. |
Tracking Brand-Associated Polarity-Bearing Topics in User Reviews (2023.tacl-1)
Copied to clipboard
| Challenge: | Existing models that infer brand polarity scores from reviews are not able to infer polarities directly. |
| Approach: | They propose a dynamic Brand-Topic Model which detects and tracks brand-associated sentiment scores and polarity-bearing topics from product reviews organized in temporally ordered time intervals. |
| Outcome: | The proposed model outperforms competitive models on a MakeupAlley and hotel review datasets. |
AutoMonitor-Bench: Evaluating the Reliability of LLM-Based Misbehavior Monitor (2026.findings-acl)
Copied to clipboard
| Challenge: | AutoMonitor-Bench evaluates the reliability of LLM-based misbehavior monitors across diverse tasks and failure modes. |
| Approach: | They introduce AutoMonitor-Bench, a benchmark designed to evaluate misbehavior monitors across diverse tasks and failure modes. |
| Outcome: | The new benchmark evaluates the reliability of LLM-based misbehavior monitors across tasks and failure modes. |
Position Bias Mitigation: A Knowledge-Aware Graph Model for Emotion Cause Extraction (2021.acl-long)
Copied to clipboard
| Challenge: | Existing models for ECE tend to explore relative position information and suffer from the dataset bias. |
| Approach: | They propose to generate adversarial examples where relative position is no longer indicative feature of cause clauses to address the dataset bias. |
| Outcome: | The proposed method performs on par with existing state-of-the-art methods on the original ECE dataset and is more robust against adversarial attacks compared to existing models. |
Beyond Prompting: An Efficient Embedding Framework for Open-Domain Question Answering (2025.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) have recently pushed open-domain question answering (ODQA) to new heights. |
| Approach: | They propose an embedding-level framework that enhances both the retriever and the reader by reordering query representations via lightweight linear layers under an unsupervised contrastive learning objective. |
| Outcome: | The proposed framework outperforms baselines in accuracy and efficiency across three open-source LLMs, three retrieval methods, and four ODQA benchmarks. |
Encourage or Inhibit Monosemanticity? Revisit Monosemanticity from a Feature Decorrelation Perspective (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent studies focus on monosemanticity on its basic units. |
| Approach: | They propose to revisit monosemanticity from the feature decorrelation perspective and advocate for its encouragement. |
| Outcome: | The proposed method improves representation diversity and activation sparsity and improves preference alignment performance. |
GraphMind: Interactive Novelty Assessment System for Accelerating Scientific Discovery (2025.emnlp-demos)
Copied to clipboard
| Challenge: | Existing approaches to literature analysis lack transparency and information retrieval module. |
| Approach: | GraphMind is an easy-to-use interactive web tool designed to assist users in evaluating novelty of scientific papers or drafted ideas. |
| Outcome: | GraphMind enables users to capture the main structure of a scientific paper, explore related ideas through various perspectives, and assess novelty via providing verifiable contextual insights. |
Beyond Cross-Modal Alignment: Measuring and Leveraging Modality Gap in Vision-Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | a recent study shows that vision-language models have modality gaps that persist even in well-aligned models. |
| Approach: | They propose a modality-dominance score to measure and leverage modality gaps . they propose automatic interpretability metrics to evaluate these features in a scalable manner . |
| Outcome: | The proposed framework allows for training-free probing and editing methods for understanding model perception across genders and generating adversarial examples. |
LexicalAT: Lexical-Based Adversarial Reinforcement Training for Robust Sentiment Classification (D19-1)
Copied to clipboard
| Challenge: | Existing text classification models are fragile and sensitive to simple perturbations. |
| Approach: | They propose a generator-classifier adversarial training approach to improve classification models . they use a large-scale lexical knowledge base to generate attacking examples . |
| Outcome: | The proposed approach outperforms strong baselines and reduces test errors on neural networks. |
Weak Reward Model Transforms Generative Models into Robust Causal Event Extraction Systems (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing evaluation metrics that reflect the performance of causal event extraction tasks are poorly reflecting the inherent ambiguity of cause and effect boundaries. |
| Approach: | They propose to use a weak-to-strong supervision method to train an evaluation model while still achieving high performance in training an RL model. |
| Outcome: | The proposed method achieves high agreement with human-annotated data while still achieving high performance in training an RL model. |
Addressing Order Sensitivity of In-Context Demonstration Examples in Causal Language Models (2024.findings-acl)
Copied to clipboard
| Challenge: | In-context learning is a popular paradigm in natural language processing, but its performance can be significantly influenced by the order of in-concept demonstration examples. |
| Approach: | They propose an unsupervised fine-tuning method to reduce the sensitivity of causal language models to the order of in-context demonstration examples. |
| Outcome: | The proposed method reduces the sensitivity of CausalLMs to the order of in-context examples and exhibits robust generalizability. |
Mirror: Multiple-perspective Self-Reflection Method for Knowledge-rich Reasoning (2024.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) struggle with knowledge-rich problems without external resources. |
| Approach: | They propose a Multiple-perspective self-reflection method that allows LLMs to reflect from multiple-perceptive clues, achieved through a heuristic interaction between a Navigator and a Reasoner. |
| Outcome: | The proposed method is superior to other self-reflection methods on five reasoning datasets. |
Drift: Enhancing LLM Faithfulness in Rationale Generation via Dual-Reward Probabilistic Inference (2025.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to improving LLM faithfulness rely on superficial calibration methods or costly retraining. |
| Approach: | They propose a probabilistic inference paradigm that leverages task-specific and lookahead rewards to ensure that LLM-generated rationales are more faithful to model decisions. |
| Outcome: | The proposed model improves both accuracy and faithfulness of Large Language Models (LLMs) on three reasoning tasks. |
Distinguishability Calibration to In-Context Learning (2023.findings-eacl)
Copied to clipboard
| Challenge: | Recent studies have shown that pre-trained language models generate similar output embeddings which makes it difficult to discriminate for the prompt-based classifier. |
| Approach: | They propose a calibration method which rotates the embedding feature into a new metric space and adapts the ratio of each dimension to a uniform distribution. |
| Outcome: | The proposed method improves the distinguishability of learning embeddings on three datasets under various settings. |