paper:arxiv-2309-11998LMSYS-Chat-1M: A large-scale real-world LLM conversation dataset
Datasets (1)
- LMSYS-CHAT-1MChat dataset (n=18,777 sampled) used to measure how much persona space PCs explain overall activation variance and to calibrate steering norms
Related work— refs + corpus + external arXiv
Cited / in-corpus / arXiv badges show which signals surfaced each row. Multi-source rows weighted higher.
- "You tell me": A Dataset of GPT-4-Based Behaviour Change Support ConversationsSelina Meyer and David Elsweiler2026≈ 77%
- LLMs and their Limited Theory of Mind: Evaluating Mental State Annotations in Situated DialogueMatthias Scheutz Katharine Kowalyshyn2025≈ 75%
- A Survey of Large Language ModelsKun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, Yifan Du, Chen Yang, Yushuo Chen, Zhipeng Chen, Jinhao Jiang, Ruiyang Ren, Yifan Li, Xinyu Tang, Zikang Liu, Peiyu Liu, Jian-Yun Nie and Ji-Rong Wen Wayne Xin Zhao2026≈ 75%
- Advances in LLMs with Focus on Reasoning, Adaptability, Efficiency and EthicsMuhammad Zaeem Khan, Aleesha Zainab, Saleha Jamshed, Sadia Ahmad, Kaynat Khatib, Faria Bibi, and Abdul Rehman Asifullah Khan2026≈ 74%
- CoMMET: To What Extent Can LLMs Perform Theory of Mind Tasks?Weifeng Jiang, Chengwei Qin, Cheston Tan Ruirui Chen2026≈ 74%
- ≈ 74%
- Beyond Self-Talk: A Communication-Centric Survey of LLM-Based Multi-Agent SystemsZhibo Zhou, Litian Zhang, Lian Zhang, Ziyi Zhou, Dezhuang Miao, Zhoujun Li, Chaozhuo Li, Xiaoming Zhang Bingyu Yan2025≈ 74%
- Can LLMs Help Decentralized Dispute Arbitration? A Case Study of UMA-Resolved Markets on PolymarketJuncen Zhou, Junjie Huang Junhao Wen2026≈ 74%
- Evaluating Large Language Models with PsychometricsYue Huang, Hongyi Wang, Ying Cheng, Xiangliang Zhang, James Zou, Lichao Sun Yuan Li2025≈ 73%
- ≈ 73%
- LLMs-Healthcare : Current Applications and Challenges of Large Language Models in various Medical SpecialtiesAwais Ahmed, Summaya Mumtaz Ummara Mumtaz2026≈ 73%
- PragWorld: A Benchmark Evaluating LLMs' Local World Model under Minimal Linguistic Alterations and Conversational DynamicsAryan Bibhuti, Atharva Naik, Martin Tutek, Somak Aditya Sachin Vashistha2025≈ 73%
- ≈ 73%
- Enhancing Conversational Agents with Theory of Mind: Aligning Beliefs, Desires, and Intentions for Human-Like InteractionDevin Yuncheng Hua, Hao Xue, Flora Salim Mehdi Jafari2025≈ 73%
- LMAgent: A Large-scale Multimodal Agents Society for Multi-user SimulationWu Liu, Xiaoyan Gu, Yong Rui, Xiaodong He and Yongdong Zhang Yijun Liu2024≈ 73%
- Quantitative Introspection in Language Models: Tracking Emotive States Across Conversationin corpus2026≈ 70%
- The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasetsin corpus2023≈ 69%
- ≈ 68%
- Monitoring and Discovering Reward Hacking with Internal Representations during LLM Evaluationsin corpus2026≈ 68%
- ≈ 67%
- ≈ 67%
- ≈ 67%
- Testing the Limits of Truth Directions in LLMsin corpus2026≈ 67%
- ≈ 67%
- ≈ 67%
- Psychological Steering of Large Language Modelsin corpus2026≈ 66%
- ≈ 66%
- Evaluating Language Model Character Traitsin corpus2024≈ 66%
Similar preprints — Semantic Scholar
Cited by (5)
- Understanding and Steering Llama 3 with Sparse Autoencoders - Goodfire
A sparse autoencoder trained on layer-19 residual-stream activations of Llama-3-8B-Instruct, using LMSYS-Chat-1M, yields features that can be labeled automatically, ranked by causal relevance, and use
- Turn-Averaged SAEs for Feature Discovery and Long-Context Attribution
Training a sparse autoencoder on the mean residual-stream activation of a conversation turn, rather than on individual tokens, yields features that describe a turn's high-level properties more complet
- Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
A single LLM trained to read injected residual-stream activations and answer arbitrary natural-language questions about them can audit fine-tuned models it never saw during training. Activation Oracle
- Persona Features Control Emergent Misalignment
Fine-tuning GPT-4o on narrowly incorrect datasets—spanning nine domains including health, legal, and automotive advice—generalizes into broadly malicious behavior across unrelated prompts, confirming
- Probe-Based Data Attribution: Surfacing and Mitigating Undesirable Behaviors in LLM Post-Training
Probe-based data attribution, introduced here as a method for surfacing and mitigating undesirable post-training behaviors, reduces harmful compliance in OLMo 2 7B by 63% through datapoint filtering a