InternLM 开源 SIM-CoT：为隐式思维链引入步骤级监督，论文被 ICLR 2026 接收
GitHub - InternLM/SIM-CoT: [ICLR 2026] An official implementation of "SIM-CoT: Supervised Implicit Chain-of-Thought" · GitHub
InternLM 发布 SIM-CoT（Supervised Implicit Chain-of-Thought）官方实现，通过训练时附加轻量辅助解码器，将每个隐式 latent token 对齐到对应推理步骤，解决隐式 CoT 随 token 数增加出现的语义同质化与训练崩溃问题，该辅助解码器在推理时移除，不增加额外推理开销。在 GPT-2 上，SIM-CoT 比监督 CoT 高 2.1%、比 Coconut 高 8.2%、比 CODI 高 4.3%；在 LLaMA 1B/3B/8B 上提升 1.5% 至 9.0%，并在 8–16 个隐式 token 下保持稳定。论文已被 ICLR 2026 接收，代码与论文已发布。
正文
当前显示原文，中文全文译文尚未就绪。
InternLM; SIM-CoT
Repository navigation
NotificationsYou must be signed in to change notification settings; Fork16 (16); Star218 (218)You must be signed in to star a repository
[ICLR 2026] An official implementation of "SIM-CoT: Supervised Implicit Chain-of-Thought"
Repository files navigation
README; Apache-2.0 license
Introduction
🌈 SIM-CoT (Supervised Implicit Chain-of-Thought) is a training framework for implicit reasoning that makes latent (implicit) CoT stable, scalable, and interpretable.
While implicit CoT can greatly reduce inference-time token cost compared to explicit chain-of-thought, prior approaches often suffer from latent instability when scaling the number of implicit tokens—leading to semantic homogenization, operator information loss, and even training collapse.
SIM-CoT addresses this by introducing step-level supervision for implicit latents. During training, we attach a lightweight auxiliary decoder to align each implicit latent token with a corresponding reasoning step, enforcing structured semantics in the latent space and improving optimization stability. Importantly, the auxiliary decoder is removed at inference time, so SIM-CoT preserves the efficiency advantages of implicit reasoning without adding any extra inference overhead.
💡 Highlights
🔥 Latent Instability in Implicit CoT: We systematically analyze the limitations of implicit Chain-of-Thought methods and reveal a latent instability issue—as the number of implicit tokens increases, models tend to collapse into homogeneous latent states that lose operator semantics; 🔥 Step-Level Supervision with SIM-CoT: We propose Supervised IMplicit-CoT (SIM-CoT), a plug-and-play module that introduces step-level supervision via an auxiliary decoder. This stabilizes optimization, prevents collapse, and ensures that latent tokens capture meaningful reasoning steps; 🔥 Strong and Consistent Performance: SIM-CoT consistently outperforms both explicit and implicit baselines. On GPT-2, it exceeds supervised CoT by +2.1%, Coconut by +8.2%, and CODI by +4.3%. Across larger LLaMA models (1B/3B/8B), it delivers +1.5% to +9.0% gains, and remains stable even with 8–16 implicit tokens, where prior methods collapse; 🔥 Efficiency and Interpretability: SIM-CoT adds no extra inference cost since the auxiliary decoder is discarded after training. It also provides interpretability, allowing each latent token to be decoded into a human-readable reasoning step
[2026/1/26] 🎉 Our paper is accepted to ICLR 2026!
[2025/9/24] Code and Paper are released!
👨💻 Todo
3. Training with Coconut + SIM-CoT
Step 1: Train the Coconut baseline
Step 2: Continue training with SIM-CoT
Select a checkpoint that has been expanded to predefined implicit tokens, then continue training with SIM-CoT:
4. Evaluation with Coconut + SIM-CoT
5. Training with CODI + SIM-CoT
6. Evaluation with CODI + SIM-CoT
✒️ Citation
If you find our work helpful for your research, please consider giving a star ⭐ and citation 📝
❤️ Acknowledgments
Coconut: The codebase we built upon. Thanks for their wonderful work; CODI: Our work is based on this codebase; we are grateful for their valuable contribution; LLaMA series: The amazing open-sourced large language model!; GPT2: An impressive open-source large language model!
[ICLR 2026] An official implementation of "SIM-CoT: Supervised Implicit Chain-of-Thought"