Structural Process Supervision for Latent Chain-of-Thought Reasoning
cs.AI
Submitted: 2026-09-09
Updated: 2026-09-28
License: http://creativecommons.org/licenses/by/4.0/
The gist: Latent reasoning approaches enhance token-level efficiency and robustness by replacing verbose, explicit chain-of-thought (CoT) tokens with compact continuous-space embeddings.
Terminology
Abstract
Latent reasoning approaches enhance token-level efficiency and robustness by replacing verbose, explicit chain-of-thought (CoT) tokens with compact continuous-space embeddings. However, existing methods lack direct process supervision over these latent embeddings, which often leads to representation collapse and uneven information distribution. To address this, we propose Prototype-Mediated Process Supervision (PMPS), which introduces learnable reasoning prototypes as semantic anchors to provide structural process-level supervision for latent reasoning. PMPS projects latent embeddings and explicit CoT embeddings into a shared prototype space, achieving many-to-many soft alignment between unequal-length representations through prototype assignment. Meanwhile, we introduce a Progressive Sequential Alignment (PSA) module to further guide training: positional priors initially encourage sequential alignment structure, then gradually relax to permit adaptive matching. Experimental results show that PMPS compresses output token length to under 50% of explicit CoT on GSM8K-Aug. Compared to leading baseline SIM-CoT, our method achieves average accuracy gains of 2.08% across different model families. On GPT-2, PMPS even surpasses CoT-SFT. On larger models and a more challenging task, PMPS consistently attains the highest accuracy among all latent reasoning methods with comparable output length.
Sources
- GPT-4 Technical Report
- Reasoning Beyond Language: A Comprehensive Survey on Latent Chain-of-Thought Reasoning
- Training Verifiers to Solve Math Word Problems
- Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities
- LLM Latent Reasoning as Chain of Superposition
- Implicit Chain of Thought Reasoning via Knowledge Distillation
- From Explicit CoT to Implicit CoT: Learning to Internalize CoT Step by Step
- The Llama 3 Herd of Models
- Training Large Language Models to Reason in a Continuous Latent Space
- Latent Visual Reasoning
- Implicit Reasoning in Large Language Models: A Comprehensive Survey
- Prototypical Contrastive Learning of Unsupervised Representations
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- SIM-CoT: Supervised Implicit Chain-of-Thought
- Qwen2.5 Technical Report
- Qwen3 Technical Report
- GLM-5: from Vibe Coding to Agentic Engineering
- A Survey on Latent Reasoning
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
- VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection