🧑‍🔬 Yao Hu

🏛️ Affiliations

🏢Xiaohongshu (Little Red Book)Shanghai · China2026 · 5 paperscurrent

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. PyramidInfer: Pyramid KV Cache Compression for High-throughput LLM Inference (2024 · 91 citations)
   Serving LLMs at scale is limited by GPU memory, and the KV cache often takes far more memory than the model weights.
2. Towards the Law of Capacity Gap in Distilling Language Models (2025 · 27 citations)
   Knowledge distillation compresses a large teacher language model into a smaller student. However, a bigger teacher does not always produce a better student. This "curse of capacity gap" means the best teacher size for a given student is usually found by distilling from many teachers, which is too costly for LLMs. The…
3. LLM-Powered Benchmark Factory: Reliable, Generic, and Efficient (2026 · 3 citations)
   With tens of thousands of new LLMs appearing each season, practitioners need custom benchmarks to choose models for their tasks.
4. Anti-Length Shift: Dynamic Outlier Truncation for Training Efficient Reasoning Models (2026 · 1 citation)
   Reasoning models trained with RLVR tend to write very long chains of thought even for easy questions, which makes deployment expensive.
5. Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors (2026 · 1 citation)
   Small tool-calling LLMs handle multi-turn tasks poorly once a tool call fails. Instead of reading the error message and adjusting, they often repeat the same invalid call until the dialogue breaks down. On BFCL v4 Multi-Turn, Qwen3-8B recovers in only about 20% of such cases, while Claude Sonnet 4 recovers in over…
6. SPARD: Self-Paced Curriculum for RL Alignment via Integrating Reward Dynamics and Data Utility (2026 · 1 citation)
   LLM post-training increasingly targets open-ended tasks, where rewards come from several criteria scored by an LLM judge.
7. Leveraging Outline-Optimized Generative Interactions and Critique for Self-Refining Outlines with Reinforcement Learning (2026)
   Small language models struggle to produce hierarchical outlines for Wikipedia-style expository articles. Their outlines tend to skip levels, omit important branches and distribute depth poorly. Existing remedies such as STORM, OmniThink or RAPID rely on costly multi-turn LLM interaction or fixed refinement…
8. Pause or Fabricate? Training Language Models for Grounded Reasoning (2026)
   The paper examines a failure mode the authors call ungrounded reasoning. When a question lacks a needed fact, a language model often notices the gap early but then invents a value and reasons confidently to a wrong answer. The authors argue that the cause is weak awareness of when valid inference is possible, not…

🤝 Co-authors here

12 of 34 co-authors with a page, the most papers together first.

• Yan Gao ×4
• Chengqiang Lu ×2
• Daoxin Zhang ×2
• Weiming Lu ×2
• Yiwu ×2
• Yongliang Shen ×2
• Bin Liang
• Chen Zhang
• Dawei Song
• Defu Lian
• Fei Zhao
• Hai Zhao

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) ×6: ACL 2026 ×5, ACL 2025
• Findings of the Association for Computational Linguistics ×2: Findings of the Association for Computational Linguistics 2026, Findings of the Association for Computational Linguistics 2024

🔗 Cited by

Papers of our library cite their works 89 times (of the 149,439 papers whose bibliographies we read); 15 citing papers with a page here →.

Updated 1 November 2026

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