A Sense of Proprioception in Language Models
cs.LG, cs.CL
Submitted: 2026-10-12
Updated: 2026-10-12
Comments: 28 pages, 22 figures, preprint
Code: https://github.com/vllm-project/vllm
License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
Sources
- Evidence for Limited Metacognition in LLMs
- LFM2 Technical Report
- Large Language Models Report Subjective Experience Under Self-Referential Processing
- Tell me about yourself: LLMs are aware of their learned behaviors
- Principles for Responsible AI Consciousness Research
- Consciousness in Artificial Intelligence: Insights from the Science of Consciousness
- The Societal Response to Potentially Sentient AI
- Could a Large Language Model be Conscious?
- Self-Cognition in Large Language Models: An Exploratory Study
- Persona Vectors: Monitoring and Controlling Character Traits in Language Models
- Does It Make Sense to Speak of Introspection in Large Language Models?
- GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models
- Introspective Coupling: Self-Explanation Training Tracks Behavioral Change Despite Fixed Supervision
- Introspection Fine-Tuning (IFT): Training Small LLMs to Introspect
- Language Models Are Capable of Metacognitive Monitoring and Control of Their Internal Activations
- Mixtral of Experts
- Robot Body Schema Learning from Full-body Extero/Proprioception Sensors
- Language Models (Mostly) Know What They Know
- Revealing Hidden Model Behaviors with Task-Specific Self-Reports
- Emergent Introspection in AI is Content-Agnostic
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks