🧑‍🔬 Baolong Bi

🏛️ Affiliations

🏛️Chinese Academy of SciencesBeijing · China2024–2026 · 37 paperscurrent🔬Institute of Computing TechnologyBeijing · China2023–2026 · 37 papersEmployment since 2023current🎓University of Chinese Academy of SciencesBeijing · China2024–2026 · 17 paperscurrent🏢MeituanBeijing · China2026 · 1 papercurrent🎓Tsinghua UniversityBeijing · 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. Neuron-Level Sequential Editing for Large Language Models (2025 · 15 citations)
   Large language models store facts that go out of date, and full retraining is expensive. Parameter-editing methods such as ROME and MEMIT work well for a single round of edits. When edits are applied one after another, however, they tend to lose earlier edits ("model forgetting") and can collapse into incoherent…
2. Can Graph Descriptive Order Affect Solving Graph Problems with LLMs? (2025 · 10 citations)
   Studies of LLM graph reasoning usually turn a graph into text by listing its edges in an arbitrary sequence.
3. Towards Fully Exploiting LLM Internal States to Enhance Knowledge Boundary Perception (2025 · 9 citations)
   LLMs often answer confidently even when they are wrong, which shows they do not reliably sense where their knowledge ends.
4. Gated Differentiable Working Memory for Long-Context Language Modeling (2026 · 4 citations)
   Long inputs remain difficult for transformer LLMs. Attention spreads thin over thousands of tokens, and evidence placed mid-document is often overlooked. Test-time training methods such as qTTT address this by briefly fine-tuning "fast weights" on the given context. However, they sample positions uniformly, so they…
5. HiddenGuard: Fine-Grained Safe Generation with Specialized Representation Router (2026 · 4 citations)
   Safety alignment in LLMs usually works as a yes/no gate: a prompt is either answered or refused. This tends to block harmless requests, such as a "killer slideshow," and to miss subtle or context-dependent harm. The authors propose HiddenGuard, which moderates output token by token instead of refusing whole answers.
6. a1: Steep Test-time Scaling Law via Environment Augmented Generation (2026 · 2 citations)
   LLMs often fail at long multi-step reasoning. They usually generate a whole solution in one pass with no intermediate checks, so an early mistake carries through to the final answer. The authors propose Environment Augmented Generation (EAG), which casts reasoning as a Markov Decision Process. While generating, the…

🤝 Co-authors here

12 of 15 co-authors with a page, the most papers together first.

• Xueqi Cheng ×5
• Lingrui Mei ×4
• Shenghua Liu ×4
• Yiwei Wang ×4
• Yuyao Ge ×3
• Jiafeng Guo ×2
• Jun Wan ×2
• An Zhang
• Junfeng Fang
• Keping Bi
• Ruibin Yuan
• Shiyu Ni

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) ×5: ACL 2026 ×2, ACL 2025 ×3
• Findings of the Association for Computational Linguistics: Findings of the Association for Computational Linguistics 2026

🔗 Cited by

Papers of our library cite their works 35 times (of the 149,439 papers whose bibliographies we read); 11 citing papers with a page here →.

Updated 31 October 2026

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