Papers by Fuli Feng

Decoding Matters: Addressing Amplification Bias and Homogeneity Issue in Recommendations for Large Language Models (2024.emnlp-main)

Challenge:	Existing studies construct attack prompts via manual or automatic methods, but these methods have limitations on cost and quality.
Approach:	They propose an attack framework to instruct LLMs to mimic human-generated prompts through in-context learning and a defense framework that fine-tunes victim LLM's through iterative interactions with the attack framework.
Outcome:	The proposed approach is based on experiments on different LLMs to evaluate their effectiveness against red teaming attacks.

Copied to clipboard

Keqin Bao, Jizhi Zhang, Yang Zhang, Xinyue Huo, Chong Chen, Fuli Feng

Dual-Phase Accelerated Prompt Optimization (2024.findings-emnlp)

Challenge:	Existing approaches to adapt Large Language Models (LLMs) for recommendation encounter significant challenges such as amplification bias and homogeneity.
Approach:	They propose a new decoding approach called Debiasing-Diversifying Decoding (D3) that disables length normalization for ghost tokens to alleviate amplification bias and incorporates a text-free assistant model to encourage tokens less frequently generated by LLMs for counteracting recommendation homogeneity.
Outcome:	Extensive experiments on real-world datasets demonstrate the proposed approach’s effectiveness in enhancing accuracy and diversity.

Copied to clipboard

Muchen Yang, Moxin Li, Yongle Li, Zijun Chen, Chongming Gao, Junqi Zhang, Yangyang Li, Fuli Feng

Latent Inter-User Difference Modeling for LLM Personalization (2025.emnlp-main)

Challenge:	Existing methods for prompt optimization make light of the importance of high-quality initialization and the identification of effective directions.
Approach:	They propose a method which uses a meta-instruction to generate high-quality initial prompts and iteratively optimize them at the sentence level.
Outcome:	The proposed method achieves consistent accuracy gain over baselines with less than five optimization steps.

Copied to clipboard

Yilun Qiu, Tianhao Shi, Xiaoyan Zhao, Fengbin Zhu, Yang Zhang, Fuli Feng

Efficient Multi-Agent System Training with Data Influence-Oriented Tree Search (2026.acl-long)

Challenge:	Large language models (LLMs) are increasingly integrated into users’ daily lives, leading to a growing demand for personalized outputs.
Approach:	They propose a framework that models inter-user differences in the latent space instead of relying on language-based prompts.
Outcome:	The proposed framework outperforms baseline methods on personalized review generation.

Copied to clipboard

Wentao Shi, Zichun Yu, Fuli Feng, Xiangnan He, Chenyan Xiong

Assistant-Guided Mitigation of Teacher Preference Bias in LLM-as-a-Judge (2025.findings-emnlp)

Challenge:	Large Language Model (LLM) based multi-agent systems (MAS) have high potential for tackling complex tasks through collaborative intelligence.
Approach:	They propose a framework that incorporates influence scores to guide tree search and data selection in data synthesis.
Outcome:	The proposed framework incorporates influence scores to guide tree search and data selection in data synthesis.

Copied to clipboard

Zhuo Liu, Moxin Li, Xun Deng, Qifan Wang, Fuli Feng

Challenge:	LLM-as-a-Judge uses large language models to evaluate the quality of LLM generated responses, but training proxy judge models using evaluation data generated by powerful teacher models introduces a critical yet previously overlooked issue: teacher preference bias.
Approach:	They propose a new setting that incorporates an additional assistant model, which is not biased toward the teacher model’s responses, to complement the training data.
Outcome:	The proposed model reduces teacher preference bias while maintaining strong performance across six evaluation benchmarks.

Hypothetical Training for Robust Machine Reading Comprehension of Tabular Context (2023.findings-acl)

Copied to clipboard

Moxin Li, Wenjie Wang, Fuli Feng, Hanwang Zhang, Qifan Wang, Tat-Seng Chua

PERM: Psychology-grounded Empathetic Reward Modeling for Large Language Models (2026.findings-acl)

Challenge:	Counterfactual training is expensive because of the complexity of tabular data.
Approach:	They propose a hypothetical training framework that uses paired examples with different hypothetical questions to supervise the direction of model gradient towards the counterfactual answer change.
Outcome:	The proposed framework improves on tabular MRC datasets.

Copied to clipboard

Chengbing Wang, Wuqiang Zheng, Yang Zhang, Fengbin Zhu, Junyi Cheng, Yi Xie, Wenjie Wang, Fuli Feng

TAT-QA: A Question Answering Benchmark on a Hybrid of Tabular and Textual Content in Finance (2021.acl-long)

Challenge:	Existing reward models evaluate empathy from a single perspective, overlooking bidirectional interaction nature of empathy.
Approach:	They propose a reward model that evaluates empathy from a single perspective . they propose PERM to integrate a bystander perspective to monitor overall interaction quality .
Outcome:	a new reward model outperforms state-of-the-art models on an emotional intelligence benchmark and an industrial daily conversation dataset.

Copied to clipboard

Fengbin Zhu, Wenqiang Lei, Youcheng Huang, Chao Wang, Shuo Zhang, Jiancheng Lv, Fuli Feng, Tat-Seng Chua

Structured Data in NLP

Challenge:	Existing QA systems focus on unstructured text, structured knowledge base, or semi-structured tables.
Approach:	They propose a large-scale question answering model based on financial reports . numerical reasoning is usually required to infer the answer .
Outcome:	The proposed model achieves 58.0% inF1, an 11.1% increase over the baseline model, but still lags behind the best human model.

Examining False Positives under Inference Scaling for Mathematical Reasoning (2025.emnlp-main)

Copied to clipboard

Yu Wang, Nan Yang, Liang Wang, Furu Wei, Fuli Feng

Leveraging Unpaired Feedback for Long-Term LLM-based Recommendation Tuning (2025.findings-emnlp)

Challenge:	Recent advances in language models have led to significant improvements in mathematical reasoning across benchmarks.
Approach:	They analyze the prevalence of false positives in language models by using heuristic evaluation methods . they find that false positive models produce correct final answers but with flawed deduction paths .
Outcome:	The proposed model performance improvements are based on the proposed model and its evaluation metrics.

Copied to clipboard

Jizhi Zhang, Chongming Gao, Wentao Shi, Xin Chen, Jingang Wang, Xunliang Cai, Fuli Feng

Challenge:	a recent study highlights unpaired feedback as a key challenge for long-term LLM-based recommenders . unpaired user feedback is crucial for improving LLMs in dynamic user environments, authors say .
Approach:	They propose a framework that incorporates unpaired feedback into LLMs to improve long-term recommendation performance.
Outcome:	The proposed framework improves long-term recommendation performance by incorporating unpaired feedback without requiring paired supervision.

Counterfactual Debating with Preset Stances for Hallucination Elimination of LLMs (2025.coling-main)

Copied to clipboard

Yi Fang, Moxin Li, Wenjie Wang, Lin Hui, Fuli Feng

MUSTIE: Multimodal Structural Transformer for Web Information Extraction (2023.acl-long)

Challenge:	Existing solutions to alleviate hallucination have considered utilizing LLMs’ inherent reasoning abilities to alleviating hallucinism, such as self-correction and diverse sampling methods.
Approach:	They propose a counterfactual multi-agent debate framework that predetermines LLMs' stances to override their inherent biases for answer inspection.
Outcome:	Extensive experiments on four datasets of three tasks demonstrate the superiority of the proposed framework over existing methods.

Copied to clipboard

Qifan Wang, Jingang Wang, Xiaojun Quan, Fuli Feng, Zenglin Xu, Shaoliang Nie, Sinong Wang, Madian Khabsa, Hamed Firooz, Dongfang Liu

Multimodality

Challenge:	Recent sequential modeling approaches focus on extracting information from textual sources while ignoring rich information from other modalities such as image and web layout.
Approach:	They propose a novel MUltimodal Structural Transformer that integrates multiple modalities for web information extraction.
Outcome:	The proposed model outperforms existing methods on WebSRC and Common Crawl benchmarks.

Controllable LLM Reasoning via Sparse Autoencoder-Based Steering (2026.acl-long)

Copied to clipboard

Yi Fang, Wenjie Wang, Mingfeng Xue, Boyi Deng, Fengli Xu, Dayiheng Liu, Fuli Feng

M2PT: Multimodal Prompt Tuning for Zero-shot Instruction Learning (2024.emnlp-main)

Challenge:	Existing methods struggle to control fine-grained reasoning strategies due to conceptual entanglement in LRMs’ hidden states.
Approach:	They propose to decompose strategy-entangled hidden states into a disentangled feature space by using Sparse Autoencoders to identify the few strategy-specific features from the vast pool of SAE features.
Outcome:	The proposed method outperforms existing methods by 15% in control effectiveness.

Copied to clipboard

Taowen Wang, Yiyang Liu, James Liang, Junhan Zhao, Yiming Cui, Yuning Mao, Shaoliang Nie, Jiahao Liu, Fuli Feng, Zenglin Xu, Cheng Han, Lifu Huang, Qifan Wang, Dongfang Liu

Unveiling Language-Specific Features in Large Language Models via Sparse Autoencoders (2025.acl-long)

Challenge:	Multimodal Large Language Models (MLLMs) exhibit remarkable performance across a wide range of domains.
Approach:	They propose a multimodal prompt tuning approach for efficient instruction tuning of MLLMs.
Outcome:	The proposed approach shows superior performance on multimodal evaluation datasets compared to state-of-the-art methods.

Copied to clipboard

Boyi Deng, Yu Wan, Baosong Yang, Yidan Zhang, Fuli Feng

ReList: A Multi-objective Reasoning Framework for Diversified Listwise Query Recommendation (2026.acl-industry)

Challenge:	Large Language Models (LLMs) exhibit impressive abilities in various domains such as text generation, instruction following, and reasoning.
Approach:	They propose a method to decompose the activations of Large Language Models into a sparse linear combination of SAE features.
Outcome:	The proposed method shows that some features are strongly related to specific languages, while others are unaffected by ablating them.

Copied to clipboard

Shuxian Bi, Chenxu Wang, Wenjie Wang, Yueqi Mou, Fuli Feng, Tang Biao, Peng Yan

MixPAVE: Mix-Prompt Tuning for Few-shot Product Attribute Value Extraction (2023.findings-acl)

Challenge:	Existing methods for related search have limited semantic redundancy and wasted retrieval quota . generative retrieval approaches lack explicit reasoning, relying on superficial click-through rate rewards .
Approach:	They propose a framework that transforms related search into a reasoning-enhanced listwise generation task.
Outcome:	Experimental results show that ReList outperforms state-of-the-art methods in query diversity and user engagement.

Copied to clipboard

Li Yang, Qifan Wang, Jingang Wang, Xiaojun Quan, Fuli Feng, Yu Chen, Madian Khabsa, Sinong Wang, Zenglin Xu, Dongfang Liu

Measuring What Makes You Unique: Difference-Aware User Modeling for Enhancing LLM Personalization (2025.findings-acl)

Challenge:	Existing methods for product attribute value extraction focus on extracting values for a set of known attributes with sufficient training data.
Approach:	They propose a prompt tuning approach to extract attributes from product information using mixed prompts.
Outcome:	The proposed approach improves on two product benchmarks and shows parameter-efficient training and avoids model overfitting.

Copied to clipboard

Yilun Qiu, Xiaoyan Zhao, Yang Zhang, Yimeng Bai, Wenjie Wang, Hong Cheng, Fuli Feng, Tat-Seng Chua

Challenge:	Extensive experiments on real-world datasets demonstrate that DPL significantly enhances LLM personalization.
Approach:	They propose a novel approach that emphasizes extracting inter-user differences to enhance LLM personalization.
Outcome:	The proposed approach extracts inter-user differences to enhance LLM personalization.

Decoding in Latent Spaces for Efficient Inference in LLM-based Recommendation (2025.findings-emnlp)

Copied to clipboard

Chengbing Wang, Yang Zhang, Zhicheng Wang, Tianhao Shi, Keqin Bao, Fuli Feng, Tat-Seng Chua

Green & Sustainable NLP

Challenge:	Light Latent-space Decoding (L2D) is an efficient and efficient latent- space decoding method.
Approach:	They propose to bypass language-space decoding by matching candidate items with LLM's internal thought representations in the latent space.
Outcome:	The proposed method is 10x faster than language-space decoding while maintaining or enhancing performance.

AJ-Bench: Benchmarking Agent-as-a-Judge for Environment-Aware Evaluation (2026.findings-acl)

Copied to clipboard

Wentao Shi, Yu Wang, Yuyang Zhao, Yuxin Chen, Fuli Feng, Xueyuan Hao, Xi Su, Qi GU, Hui Su, Xunliang Cai, Xiangnan He

Challenge:	Existing approaches to verify agent behaviors in complex environments rely on rule-based verifiers or LLM-as-a-Judge models.
Approach:	They propose a benchmark to evaluate Agent-as-a-Judge across three domains . the benchmark covers search, data systems, and graphical user interfaces - with 155 tasks and 516 trajectories .
Outcome:	The proposed benchmark outperforms existing benchmarks in search, data systems, and GUI domains while revealing open challenges in agent-based verification.

Learning to Imagine: Integrating Counterfactual Thinking in Neural Discrete Reasoning (2022.acl-long)

Copied to clipboard

Moxin Li, Fuli Feng, Hanwang Zhang, Xiangnan He, Fengbin Zhu, Tat-Seng Chua

APrompt: Attention Prompt Tuning for Efficient Adaptation of Pre-trained Language Models (2023.emnlp-main)

Challenge:	Existing NDR models suffer from large performance drop on hypothetical questions, e.g., “what the annualized rate of return would be if the revenue in 2020 was doubled”.
Approach:	They propose a learning to imagine module which can be seamlessly incorporated into NDR models to perform the imagination of unseen counterfactual.
Outcome:	The proposed model can perform the imagination of unseen counterfactuals on hypothetical questions.

Copied to clipboard

Qifan Wang, Yuning Mao, Jingang Wang, Hanchao Yu, Shaoliang Nie, Sinong Wang, Fuli Feng, Lifu Huang, Xiaojun Quan, Zenglin Xu, Dongfang Liu

HellaSwag-Pro: A Large-Scale Bilingual Benchmark for Evaluating the Robustness of LLMs in Commonsense Reasoning (2025.findings-acl)

Challenge:	Existing prompt tuning methods only introduce prompts at the input layer, limiting performance and leaving large room for improvement.
Approach:	They propose a method that involves tuning a small set of soft prompts for pre-trained language models.
Outcome:	The proposed method outperforms state-of-the-art methods with pre-trained models on the SuperGLUE benchmark.

Copied to clipboard

Xiaoyuan Li, Moxin Li, Rui Men, Yichang Zhang, Keqin Bao, Wenjie Wang, Fuli Feng, Dayiheng Liu, Junyang Lin

Robustness in NLP

Challenge:	Existing studies show that large language models are robust in commonsense reasoning . however, some variations in questions can lead to incorrect responses .
Approach:	They propose a large-scale bilingual benchmark consisting of 11,200 cases . they conduct extensive experiments on 41 representative LLMs .
Outcome:	The proposed benchmark systematically evaluates the robustness of large language models in commonsense reasoning.

MTR-Bench: A Comprehensive Benchmark for Multi-Turn Reasoning Evaluation (2026.acl-long)

Copied to clipboard

Xiaoyuan Li, Keqin Bao, Yubo Ma, Moxin Li, Wenjie Wang, Rui Men, Yichang Zhang, Fuli Feng, Dayiheng Liu

Evaluating Mathematical Reasoning of Large Language Models: A Focus on Error Identification and Correction (2024.findings-acl)

Challenge:	Recent advances in Large Language Models (LLMs) have shown promising results in complex reasoning tasks.
Approach:	They propose to use a multi-turn reasoning evaluation framework to cover multi-turn interactions with the environments of large language models.
Outcome:	The proposed framework covers diverse reasoning capabilities, fine-grained difficulty granularity, and necessitates multi-turn interactions with the environments.

Copied to clipboard

Xiaoyuan Li, Wenjie Wang, Moxin Li, Junrong Guo, Yang Zhang, Fuli Feng

Disentangling Reasoning Tokens and Boilerplate Tokens For Language Model Fine-tuning (2025.findings-acl)

Challenge:	Existing evaluations focus on problem-solving from examiner perspective, overlooking a dual perspective of examiner regarding error identification and correction.
Approach:	They propose to use an annotated dataset to evaluate large language models from the examiner perspective and to use diverse prompts to evaluate eleven representative LLMs.
Outcome:	The proposed model outperforms all models while LLaMA-2-7B has comparable abilities to closed-source models GPT-3.5 and Gemini Pro.

Copied to clipboard

Ziang Ye, Zhenru Zhang, Yang Zhang, Jianxin Ma, Junyang Lin, Fuli Feng

Consistency-Aware Online Multi-Objective Alignment for Related Search Query Generation (2025.acl-industry)

Challenge:	Existing approaches to enhance agent capabilities for Large Language Models treat all tokens equally . however, reasoning tokens versus boilerplate tokens differ in importance and learning complexity . recent research has focused on enhancing agent capabilities in large language models .
Approach:	They propose a Shuffle-Aware Discriminator (SHAD) for adaptive token discrimination . they propose SHAD method which adaptively emphasizes reasoning tokens during fine-tuning .
Outcome:	The proposed method improves performance over standard fine-tuning methods.

Copied to clipboard

Shuxian Bi, Chongming Gao, Wenjie Wang, Yueqi Mou, Chenxu Wang, Tang Biao, Peng Yan, Fuli Feng

Information Retrieval

Challenge:	Existing methods fail to reconcile click-through rate (CTR) optimization with topic expansion.
Approach:	They propose a query generation framework that aligns click-through rate and topic expansion goals through an online DPO paradigm.
Outcome:	The proposed approach achieves significant CTR gains (+2.3%) and higher human-rated query quality compared to state-of-the-art methods.

Text-like Encoding of Collaborative Information in Large Language Models for Recommendation (2024.acl-long)

Copied to clipboard

Yang Zhang, Keqin Bao, Ming Yan, Wenjie Wang, Fuli Feng, Xiangnan He

Self-Improvement Towards Pareto Optimality: Mitigating Preference Conflicts in Multi-Objective Alignment (2025.findings-acl)

Challenge:	Existing methods to adapt Large Language Models for Recommendation (LLMRec) do not represent collaborative information in a text-like format, which may not align optimally with LLMs.
Approach:	They propose a novel LLMRec method that integrates collaborative information through text-like encoding.
Outcome:	Extensive experiments show that BinLLM integrates collaborative information better with LLMs.

Copied to clipboard

Moxin Li, Yuantao Zhang, Wenjie Wang, Wentao Shi, Zhuo Liu, Fuli Feng, Tat-Seng Chua

Counterfactual Inference for Text Classification Debiasing (2021.acl-long)

Challenge:	Existing approaches to optimize large language models with human preferences suffer from preference conflicts in the data.
Approach:	They propose to construct Pareto-optimal responses to resolve preference conflicts by using a self-improving DPO framework that enables LLMs to self-generate and select Paret-optimized responses.
Outcome:	The proposed framework achieves superior Pareto Front performance over baselines on two datasets.

Copied to clipboard

Chen Qian, Fuli Feng, Lijie Wen, Chunping Ma, Pengjun Xie

Robustness in NLP

Challenge:	Existing methods to capture unintended dataset biases are expensive and require elaborate balancing strategies.
Approach:	They propose a model-agnostic text classification debiasing framework which can effectively avoid employing data manipulations or designing balancing mechanisms.
Outcome:	The proposed framework can effectively avoid data manipulations or designing balancing mechanisms to capture unintended dataset biases.

Learning to Generate Question by Asking Question: A Primal-Dual Approach with Uncommon Word Generation (2022.emnlp-main)

Copied to clipboard

Qifan Wang, Li Yang, Xiaojun Quan, Fuli Feng, Dongfang Liu, Zenglin Xu, Sinong Wang, Hao Ma

Text Generation

Challenge:	Existing automatic question generation methods focus on encoding passage and answer to generate question.
Approach:	They propose an automatic question generation approach which integrates question generation with its dual problem, question answering, into a unified primal-dual framework.
Outcome:	The proposed approach outperforms existing methods on SQuAD and HotpotQA benchmarks.

Customizing In-context Learning for Dynamic Interest Adaption in LLM-based Recommendation (2025.findings-acl)

Copied to clipboard

Keqin Bao, Ming Yan, Yang Zhang, Jizhi Zhang, Wenjie Wang, Fuli Feng, Xiangnan He

Low-Resource NLP

Challenge:	Existing Large Language Model (LLM)-based recommender systems face challenges to adapt to dynamic user interests without any model-level updates.
Approach:	They propose a framework that establishes recommendation-oriented in-context learning by structuring recent user interactions and current inputs into ICL formats.
Outcome:	The proposed model adapts to dynamic user interests without model updates without any model updates and is available online at https://anonymous.4open.science/r/RecICL-8003.

Robust Prompt Optimization for Large Language Models Against Distribution Shifts (2023.emnlp-main)

Copied to clipboard

Moxin Li, Wenjie Wang, Fuli Feng, Yixin Cao, Jizhi Zhang, Tat-Seng Chua

Counterfactual Active Learning for Out-of-Distribution Generalization (2023.acl-long)

Challenge:	Existing research has explored automatic prompt optimization methods to eliminate manual effort in identifying effective prompts for a given task.
Approach:	They propose a framework for prompt optimization that can be generalized to an unlabeled target group.
Outcome:	The proposed framework improves on target group and source group while generalizing to unlabeled target group.

Copied to clipboard

Xun Deng, Wenjie Wang, Fuli Feng, Hanwang Zhang, Xiangnan He, Yong Liao

Think Twice Before Trusting: Self-Detection for Large Language Models through Comprehensive Answer Reflection (2024.findings-emnlp)

Challenge:	Existing studies on active learning methods focus on the out-of-distribution generalization of out- of-distortion samples.
Approach:	They propose a counterfactual active learning approach that empowers active learning with counterfact thinking to bridge the seen samples with unseen cases.
Outcome:	The proposed approach outperforms existing active learning methods on public datasets with comparable IID performance.

Copied to clipboard

Moxin Li, Wenjie Wang, Fuli Feng, Fengbin Zhu, Qifan Wang, Tat-Seng Chua

EAVE: Efficient Product Attribute Value Extraction via Lightweight Sparse-layer Interaction (2024.findings-emnlp)

Challenge:	Existing approaches to self-detection only retrospectively evaluate LLM-generated answers, leading to over-trust in incorrectly generated answers.
Approach:	They propose a self-detection paradigm that considers the comprehensive answer space beyond LLM-generated answers to mitigate the over-trust in LLM generated incorrect answers.
Outcome:	The proposed framework can be integrated with existing approaches for superior self-detection.

Copied to clipboard

Li Yang, Qifan Wang, Jianfeng Chi, Jiahao Liu, Jingang Wang, Fuli Feng, Zenglin Xu, Yi Fang, Lifu Huang, Dongfang Liu

Tunable LLM-based Proactive Recommendation Agent (2025.acl-long)

Challenge:	Existing methods to extract product attribute value require multiple extractions to obtain all corresponding values.
Approach:	They propose an Efficient product Attribute Value Extraction approach using lightweight sparse-layer interaction.
Outcome:	The proposed method achieves significant efficiency gains with neutral or marginal loss in performance when the context is long and number of attributes is large.

Copied to clipboard

Mingze Wang, Chongming Gao, Wenjie Wang, Yangyang Li, Fuli Feng

Challenge:	Current methods focus on catering to existing user interests, leading to polarized recommendation distributions.
Approach:	They propose an LLM-based Actor-Critic Agent framework to cultivate latent interests through multi-step recommendations.
Outcome:	The proposed framework optimizes long-term rewards and dynamically evolves with user feedback.

Empowering Language Understanding with Counterfactual Reasoning (2021.findings-acl)

Copied to clipboard

Fuli Feng, Jizhi Zhang, Xiangnan He, Hanwang Zhang, Tat-Seng Chua

RoAST: Robustifying Language Models via Adversarial Perturbation with Selective Training (2023.findings-emnlp)

Challenge:	Existing methods for language understanding use the recognized patterns in the testing phase that are inherently different from us humans who have counterfactual thinking.
Approach:	They propose a counterfactual Reasoning Model which mimics counterfactive thinking by learning from few counterffact samples.
Outcome:	The proposed model can detect and make predictions from textual patterns . it can also detect negative sarcastic puns by comparing them with imaginations .

Copied to clipboard

Jaehyung Kim, Yuning Mao, Rui Hou, Hanchao Yu, Davis Liang, Pascale Fung, Qifan Wang, Fuli Feng, Lifu Huang, Madian Khabsa

Robustness in NLP

Challenge:	Several perspectives of robustness for pre-trained language models have been studied independently, but lacking a unified consideration in multiple perspectives.
Approach:	They propose a technique to enhance the multi-perspective robustness of LMs by introducing adversarial perturbation while the model parameters are selectively updated upon their relative importance.
Outcome:	The proposed technique improves the robustness of LMs by incorporating four perspectives on model robustness.

TROJail: Trajectory-Level Optimization for Multi-Turn Large Language Model Jailbreaks with Process Rewards (2026.acl-long)

Copied to clipboard

Xiqiao Xiong, Ouxiang Li, Zhuo Liu, Moxin Li, Wentao Shi, Fengbin Zhu, Qifan Wang, Fuli Feng