🧑‍🔬 Hao Li

🏛️ Affiliations

🔬Shanghai Artificial Intelligence LaboratoryShanghai · China2025–2026 · 3 paperscurrent🏢Alibaba GroupHangzhou · China2026 · 1 papercurrent🎓Beihang UniversityBeijing · China2026 · 1 papercurrent🎓Beijing Institute of TechnologyBeijing · China2026 · 1 papercurrent🎓Northwestern Polytechnical UniversityXi'an · China2026 · 1 papercurrent

From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine.

🧭 Fields, topics and concepts

Of its works in our library, with the number of works on each.

📅 Works by year

2020s

📄 Papers

The most cited first (OpenAlex and our library).

1. Position-Aware Tagging for Aspect Sentiment Triplet Extraction (2020 · 370 citations)
   This paper addresses Aspect Sentiment Triplet Extraction (ASTE). The task is to find each target in a sentence, its sentiment polarity, and the opinion span that explains that sentiment. The only earlier method (Peng et al., 2019) used two stages: it first labelled targets with sentiment and opinion spans, then…
2. Hello Again! LLM-powered Personalized Agent for Long-term Dialogue (2025 · 55 citations)
   Most open-domain chatbots are built for short, single-session exchanges. Real companionship, however, requires remembering past conversations and keeping track of who the user and the bot are. The authors present LD-Agent, a model-agnostic framework that splits long-term dialogue into three separately trainable…
3. VLSBench: Unveiling Visual Leakage in Multimodal Safety (2025 · 32 citations)
   Earlier work reported a puzzling result: aligning multimodal LLMs with text-only data seemed to make them about as safe as aligning them with image-text pairs.
4. LLMs know their vulnerabilities: Uncover Safety Gaps through Natural Distribution Shifts (2025 · 18 citations)
   The paper investigates a safety weakness of aligned LLMs. Seemingly harmless prompts that are semantically connected to a harmful request can steer a model into producing unsafe output. The authors call this exploitation of "natural distribution shift" and contrast it with attacks that deliberately distort the toxic…
5. PIGuard: Prompt Injection Guardrail via Mitigating Overdefense for Free (2025 · 16 citations)
   Lightweight prompt guard classifiers screen LLM inputs for prompt injection, but they often learn a shortcut.
6. Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment (2025 · 11 citations)
   Fine-tuning a safety-aligned LLM can weaken its refusal behaviour even when the training data look harmless.
7. Rethinking Jailbreak Detection of Large Vision Language Models with Representational Contrastive Scoring (2026 · 10 citations)
   Multimodal jailbreak attacks on Large Vision-Language Models (LVLMs) call for detectors that work on attack types never seen in training and that add little runtime cost.
8. LogitSpec: Accelerating Retrieval-based Speculative Decoding via Next Next Token Speculation (2026 · 6 citations)
   Speculative decoding speeds up LLM inference by drafting several tokens and having the target model verify them in one pass.
9. Be Your Own Red Teamer: Safety Alignment via Self-Play and Reflective Experience Replay (2026 · 1 citation)
   LLM safety training usually depends on fixed red-teaming resources, such as static defense prompts or pre-collected jailbreak datasets.
10. Data Mixing Agent: Learning to Re-weight Domains for Continual Pre-training (2026 · 1 citation)
    Continual pre-training of an LLM on a narrow target field, such as math, often causes catastrophic forgetting of general skills.
11. AutoTaskEval: Towards Domain-Specific and Fine-Grained Evaluation for LLMs (2026)
    Static, manually built benchmarks are expensive, cover few tasks, and lose discriminative power as LLMs improve.

All works

12 works of theirs in our library, the most cited first; 11 with a page here, the others link their original until we summarise them.

1. Position-Aware Tagging for Aspect Sentiment Triplet ExtractionConference on Empirical Methods in Natural Language Processing (EMNLP)Lu Xu, Hao Li, Wei Lu, et al.2020 · 370 citations
2. Hello Again! LLM-powered Personalized Agent for Long-term DialogueConference of the North American Chapter of the Association for Computational Linguistics (NAACL)Hao Li, Chenghao Yang, An Zhang, et al.2025 · 55 citations
3. VLSBench: Unveiling Visual Leakage in Multimodal SafetyAnnual Meeting of the Association for Computational Linguistics (ACL)Xuhao Hu, Dongrui Liu, Hao Li, et al.2025 · 32 citations
4. LLMs know their vulnerabilities: Uncover Safety Gaps through Natural Distribution ShiftsAnnual Meeting of the Association for Computational Linguistics (ACL)Qibing Ren, Hao Li, Dongrui Liu, et al.2025 · 18 citations
5. PIGuard: Prompt Injection Guardrail via Mitigating Overdefense for FreeAnnual Meeting of the Association for Computational Linguistics (ACL)Hao Li, Xiaogeng Liu, Ning Zhang, et al.2025 · 16 citations
6. Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety AlignmentConference on Empirical Methods in Natural Language Processing (EMNLP)Hao Li, Lijun Li, Zhenghao Lu, et al.2025 · 11 citations
7. Rethinking Jailbreak Detection of Large Vision Language Models with Representational Contrastive ScoringAnnual Meeting of the Association for Computational Linguistics (ACL)Peichun Hua, Hao Li, Shanghao Shi, et al.2026 · 10 citations
8. LogitSpec: Accelerating Retrieval-based Speculative Decoding via Next Next Token SpeculationFindings of the Association for Computational LinguisticsTianyu Liu, Qitan Lv, Hao Li, et al.2026 · 6 citations
9. Be Your Own Red Teamer: Safety Alignment via Self-Play and Reflective Experience ReplayFindings of the Association for Computational LinguisticsHao Wang, Yanting Wang, Hao Li, et al.2026 · 1 citation
10. Data Mixing Agent: Learning to Re-weight Domains for Continual Pre-trainingAnnual Meeting of the Association for Computational Linguistics (ACL)Kailai Yang, Xiao Liu, Lei Ji, et al.2026 · 1 citation
11. AutoTaskEval: Towards Domain-Specific and Fine-Grained Evaluation for LLMsAnnual Meeting of the Association for Computational Linguistics (ACL)Qingqing Lyu, Linjuan Wu, Yongliang Shen, et al.2026
12. MTRouter: Cost-Aware Multi-Turn LLM Routing with History–Model Joint Embeddings ↗Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · no summary yetYiqun Zhang et al.2026

🤝 Co-authors here

12 of 40 co-authors with a page, the most papers together first.

• Jing Shao ×3
• Lei Sha ×3
• Dongrui Liu ×2
• Ning Zhang ×2
• Rui Li ×2
• An Zhang
• Chaowei Xiao
• Chenghao Yang
• Hao Wang
• Hengwei Liu
• Junchi Yan
• Kailai Yang

Venues

Where their works here were published, the most works first, with the editions of those that have a page.

• Annual Meeting of the Association for Computational Linguistics (ACL) ×7: ACL 2026 ×3, ACL 2025 ×3
• Conference on Empirical Methods in Natural Language Processing (EMNLP) ×2: EMNLP 2025, EMNLP 2020
• Findings of the Association for Computational Linguistics ×2: Findings of the Association for Computational Linguistics 2026 ×2
• Conference of the North American Chapter of the Association for Computational Linguistics (NAACL): NAACL 2025

🔗 Cited by

Papers of our library cite their works 207 times (of the 149,439 papers whose bibliographies we read); 50 citing papers with a page here →.

Updated 1 November 2026

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/authors/hao-li/