Papers by Ping Jian
Temporal Token Matters: Investigating and Interpreting the Consistency of Temporal Ordering in Large Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) exhibit notable deficiencies in temporal reasoning . phrasing changes can lead LLMs to produce inconsistent outputs . |
| Approach: | They investigate the mechanistic interpretability of temporal ordering within event temporal reasoning . they identify a sparse subset of attention heads that are causally responsible for reasoning outcomes . |
| Outcome: | The proposed model outperforms other models in a variety of tasks and is validated by intervention-based experiments. |
From Misleading Queries to Accurate Answers: A Three-Stage Fine-Tuning Method for LLMs (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods focus on correcting the output but overlook the ability of LLMs to detect and correct misleading content in the input itself. |
| Approach: | They propose a three-stage fine-tuning method that improves LLMs' ability to detect and correct misleading information in input queries. |
| Outcome: | The proposed method improves accuracy and factuality of LLM responses while also reducing hallucinations. |
Option Symbol Matters: Investigating and Mitigating Multiple-Choice Option Symbol Bias of Large Language Models (2025.naacl-long)
Copied to clipboard
| Challenge: | Multiple-Choice Question Answering (MCQA) is a widely used task in the evaluation of large language models (LLMs). |
| Approach: | They propose a tuning-free, causal effect driven debiasing method which intervenes the activations of identified components according to their causal effects. |
| Outcome: | The proposed method alleviates the aforementioned bias and improves the performance of LLMs. |
Induction Networks for Few-Shot Text Classification (D19-1)
Copied to clipboard
| Challenge: | Recent studies have used meta-learning to simulate the few-shot task . however, this sample-wise comparison may be severely disturbed by the various expressions in the same class. |
| Approach: | They propose a meta-learning-based induction network to learn a generalized class-wise representation of each class in a support set. |
| Outcome: | The proposed model outperforms existing state-of-the-art models on a sentiment and dialogue intent datasets. |
Improving Implicit Discourse Relation Recognition with Semantics Confrontation (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing methods for implicit discourse relation recognition (IDRR) are unsatisfactory for the task. |
| Approach: | They propose a method that trains PLMs through two semantics enhancers to implicitly differentiate logical and general semantics. |
| Outcome: | The proposed method exceeds baseline by 3.81% F1 score on PDTB 2.0 dataset . it infers discourse logical relations without explicit connectives, but performance remains unsatisfactory . |
Effective Integration of Text Diffusion and Pre-Trained Language Models with Linguistic Easy-First Schedule (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing noise schedules in text diffusion models do not take into account the linguistic differences among tokens, which violates the easy-first policy for text generation. |
| Approach: | They propose to use a pre-trained decoder to convert denoised embedding vectors into natural language instead of the widely used rounding operation. |
| Outcome: | The proposed model outperforms existing models on the E2E dataset and five controllable tasks on the discrete nature of text data. |
Constructing Your Model’s Value Distinction: Towards LLM Alignment with Anchor Words Tuning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a study of large language models (LLMs) shows that they can generate outputs that are honest, positive, harmless, etc. |
| Approach: | They propose a method that amplifies logits difference between positive and negative tokens . they propose to use the logits gap to generate positive and positive tokens after alignment . |
| Outcome: | The proposed method achieves effective alignment, but requires fewer computational resources compared to training-time alignment methods. |
Prompt-based Logical Semantics Enhancement for Implicit Discourse Relation Recognition (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing methods for identifying discourse relations without explicit connectives are limited by the availability of annotated data. |
| Approach: | They propose a method that injects knowledge relevant to discourse relation into pre-trained language models through prompt-based connective prediction. |
| Outcome: | The proposed method achieves outstanding performance against the current state-of-the-art models. |
How Do LLMs and VLMs Understand Viewpoint Rotation Without Vision? An Interpretability Study (2026.acl-long)
Copied to clipboard
Zhen Yang, Ping Jian, Zhongbin Guo, Zuming Zhang, Chengzhi Li, Yonghong Deng, Xinyue Zhang, Wenpeng Lu
| Challenge: | Existing studies on spatial intelligence from the perspective of visual-spatial intelligence have not explored whether visual intelligence alone is sufficient to endow models with spatial intelligence. |
| Approach: | They propose to use a linguistic perspective to investigate spatial intelligence from a theoretical perspective. |
| Outcome: | The proposed model performs poorly on the proposed dataset while human can easily achieve 100% accuracy. |
Where CoT Reasoning Commits: Entropy Traces Identify Interpretable Attention Heads (2026.findings-acl)
Copied to clipboard
| Challenge: | a growing body of work suggests a disconnect between the generated rationale and the model's actual choice. |
| Approach: | They propose a mechanism-aware framework that interprets the evolving "choice state" of attention heads during CoT generation . they identify a set of intervention targets and perform Selective Head Fine-Tuning . |
| Outcome: | The proposed framework interprets the "choice state" of attention heads during CoT generation . it detects two functional behaviors: Steadfast Heads and Wavering Heads . |
Context Tracking Network: Graph-based Context Modeling for Implicit Discourse Relation Recognition (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing models fail to fully utilize contextual information which plays an important role in interpreting sentences. |
| Approach: | They propose a graph-based Context Tracking Network to model the discourse context for IDRR. |
| Outcome: | The proposed model can integrate sentence-level and token-level contextual semantics better than existing models. |
Debate-of-Thoughts: Resolving Knowledge Conflicts in LLMs Through Internal Deliberation (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for retrieval augmented generation are based on a simplistic binary choice of relying on external contexts or memory. |
| Approach: | They propose a framework that transforms conflict resolution into an active deliberation process by incorporating contradictions as opportunities for deeper reasoning. |
| Outcome: | Experiments show that DoT outperforms state-of-the-art methods while generating transparent debate transcripts that explain its decisions. |
Intra-Correlation Encoding for Chinese Sentence Intention Matching (2020.coling-main)
Copied to clipboard
| Challenge: | Existing methods to improve sentence intention matching for Chinese text are limited due to the particularity of the text. |
| Approach: | They propose a method that combines character-granularity and word-granulularity features to perform sentence intention matching. |
| Outcome: | The proposed method can capture sentence feature information from multiple perspectives and correlation information between different levels of sentences. |
Memory or Reasoning? Explore How LLMs Compute Mixed Arithmetic Expressions (2025.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) can solve complex multi-step math reasoning problems, but their internal implementation is limited. |
| Approach: | They propose to use a "C**ausal **E**ffect **D**riven **F**ine-tuning method" to improve LLMs' reasoning ability. |
| Outcome: | The proposed method improves the model's reasoning ability by enhancing key components that are used to execute mixed arithmetic calculations. |
Can LLMs See Without Pixels? Benchmarking Spatial Intelligence from Textual Descriptions (2026.findings-acl)
Copied to clipboard
Zhongbin Guo, Zhen Yang, Yushan Li, Xinyue Zhang, Wenyu Gao, Jiacheng Wang, Chengzhi Li, Xiangrui Liu, Ping Jian
| Challenge: | Existing advances in Spatial Intelligence rely on vision-Language Models . however, a critical question remains: does spatial understanding originate from visual encoders? |
| Approach: | They propose to evaluate the SI performance of Large Language Models without pixel-level input. |
| Outcome: | The proposed benchmark challenges large language models to perform symbolic reasoning rather than visual pattern matching. |