mlsanityData to 7 Oct 2026

Toward IIT-Inspired Consciousness in LLMs: A Reward-Based Learning Framework

arXiv 2601.22786 ↗First posted 30 Jan 2026cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

The pursuit of Artificial General Intelligence (AGI) is a central goal in language model development, in which consciousness-like processing could serve as a key facilitator. While current language models are not conscious, they exhibit behaviors analogous to certain aspects of consciousness. This paper investigates the implementation of a leading theory of consciousness, Integrated Information Theory (IIT), within language models via a reward-based learning paradigm. IIT provides a formal, axiom-based mathematical framework for quantifying consciousness. Drawing inspiration from its core principles, we formulate a novel reward function that quantifies a text's causality, coherence and integration, characteristics associated with conscious processing. Empirically, it is found that optimizing for this IIT-inspired reward leads to more concise text generation. On out of domain tasks, careful tuning achieves up to a 31% reduction in output length while preserving accuracy levels comparable to the base model. In addition to primary task performance, the broader effects of this training methodology on the model's confidence calibration and test-time computational scaling is analyzed. The proposed framework offers significant practical advantages: it is conceptually simple, computationally efficient, requires no external data or auxiliary models, and leverages a general, capability-driven signal rather than task-specific heuristics. Code available at https://github.com/MH-Sameti/LLM_PostTraining.git

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Builds on 31 of its 54 references that are in the map. First posted on arXiv 30 Jan 2026.

Builds on

31 of 54 references.

References · 31

  1. 2023GPT-4 Technical Reportcited by 18,132
  2. 2017On Calibration of Modern Neural Networkscited by 4,860
  3. 2024Qwen2 Technical Reportcited by 2,076
  4. 2025Stop Overthinking: A Survey on Efficient Reasoning for Large Language Modelscited by 456
  5. 2025L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learningcited by 381
  6. 2025Demystifying Long Chain-of-Thought Reasoning in LLMscited by 323
  7. 2023A Long Way to Go: Investigating Length Correlations in RLHFcited by 285
  8. 2025O1-Pruner: Length-Harmonizing Fine-Tuning for O1-Like Reasoning Pruningcited by 278
  9. 2025TokenSkip: Controllable Chain-of-Thought Compression in LLMscited by 246
  10. 2024Token-Budget-Aware LLM Reasoningcited by 242
  11. 2025Training Language Models to Reason Efficientlycited by 242
  12. 2023Do Large Language Models Know What They Don't Know?cited by 238
19 more
  1. 2025The Unreasonable Effectiveness of Entropy Minimization in LLM Reasoningcited by 160
  2. 2025Thinkless: LLM Learns When to Thinkcited by 107
  3. 2023Consciousness in Artificial Intelligence: Insights from the Science of Consciousnesscited by 99
  4. 2023Knowledge of Knowledge: Exploring Known-Unknowns Uncertainty with Large Language Modelscited by 60
  5. 2024Following Length Constraints in Instructionscited by 45
  6. 2024Knowledge Boundary of Large Language Models: A Surveycited by 36
  7. 2025Toward Generalizable Evaluation in the LLM Era: A Survey Beyond Benchmarkscited by 26
  8. 2022Integrated information theory (IIT) 4.0: Formulating the properties of phenomenal existence in physical termscited by 21
  9. 2023Prompt-Based Length Controlled Generation with Reinforcement Learningcited by 14
  10. 2024Self-Cognition in Large Language Models: An Exploratory Studycited by 14
  11. 2024From Imitation to Introspection: Probing Self-Consciousness in Language Modelscited by 14
  12. 2024Precise Length Control in Large Language Modelscited by 13
  13. 2025SPRINT: Enabling Interleaved Planning and Parallelized Execution in Reasoning Modelscited by 10
  14. 2024A Case for AI Consciousness: Language Agents and Global Workspace Theorycited by 7
  15. 2024Dissociating Artificial Intelligence from Artificial Consciousnesscited by 7
  16. 2025Token-Efficient RL for LLM Reasoningcited by 6
  17. 2023Survey of Consciousness Theory from Computational Perspectivecited by 3
  18. 2019The Phi measure of integrated information is not well-defined for general physical systemscited by 1
  19. 2025Can "consciousness" be observed from large language model (LLM) internal states? Dissecting LLM representations obtained from Theory of Mind test with Integrated Information Theory and Span Representation analysiscited by 1

Built on by

Not cited yet.