mlsanityData to 14 Oct 2026

Co-Alignment: Rethinking Alignment as Bidirectional Human-AI Cognitive Adaptation

arXiv 2509.12179 ↗First posted 15 Sep 2025cs.AI cs.MAcited by 0

On the map: LLM alignment / RLHF / preference optimisation → Human values and AI feedback alignment

Abstract (arXiv)

Current AI alignment through RLHF follows a single directional paradigm that AI conforms to human preferences while treating human cognition as fixed. We propose a shift to co-alignment through Bidirectional Cognitive Alignment (BiCA), where humans and AI mutually adapt. BiCA uses learnable protocols, representation mapping, and KL-budget constraints for controlled co-evolution. In collaborative navigation, BiCA achieved 85.5% success versus 70.3% baseline, with 230% better mutual adaptation and 332% better protocol convergence. Emergent protocols outperformed handcrafted ones by 84%, while bidirectional adaptation unexpectedly improved safety (+23% out-of-distribution robustness). The 46% synergy improvement demonstrates optimal collaboration exists at the intersection, not union, of human and AI capabilities, validating the shift from single-directional to co-alignment paradigms.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Human values and AI feedback alignment, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 46 of its 81 references that are in the map. First posted on arXiv 15 Sep 2025.

Builds on

46 of 81 references.

Closest ancestors · 3

Same direction, newest first

  1. 2023Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision14 Dec
  2. 2023The Past, Present and Better Future of Feedback Learning in Large Language Models for Subjective Human Preferences and Values11 Oct
  3. 2022Constitutional AI: Harmlessness from AI Feedback15 Dec

Other references · 43

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2017Proximal Policy Optimization Algorithmscited by 15,215
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  4. 2017Deep reinforcement learning from human preferencescited by 3,912
  5. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,739
  6. 2023Sparks of Artificial General Intelligence: Early experiments with GPT-4cited by 2,592
  7. 2015Trust Region Policy Optimizationcited by 2,580
  8. 2022Emergent Abilities of Large Language Modelscited by 2,512
  9. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,191
  10. 2019Similarity of Neural Network Representations Revisitedcited by 1,755
  11. 2017Multi-Agent Actor-Critic for Mixed Cooperative-Competitive Environmentscited by 1,558
  12. 2023Towards Understanding Sycophancy in Language Modelscited by 1,040
31 more
  1. 2017Constrained Policy Optimizationcited by 983
  2. 2023Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedbackcited by 759
  3. 2018QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learningcited by 708
  4. 2020GANSpace: Discovering Interpretable GAN Controlscited by 614
  5. 2016Learning to Communicate with Deep Multi-Agent Reinforcement Learningcited by 575
  6. 2019Multi-Agent Reinforcement Learning: A Selective Overview of Theories and Algorithmscited by 567
  7. 2017Word Translation Without Parallel Datacited by 487
  8. 2020Artificial Intelligence, Values and Alignmentcited by 448
  9. 2020Closed-Form Factorization of Latent Semantics in GANscited by 408
  10. 2020Does the Whole Exceed its Parts? The Effect of AI Explanations on Complementary Team Performancecited by 399
  11. 2016Cooperative Inverse Reinforcement Learningcited by 300
  12. 2021AI Chains: Transparent and Controllable Human-AI Interaction by Chaining Large Language Model Promptscited by 268
  13. 2017Emergence of Grounded Compositional Language in Multi-Agent Populationscited by 263
  14. 2017Automated Curriculum Learning for Neural Networkscited by 257
  15. 2019Vision-and-Dialog Navigationcited by 250
  16. 2018Touchdown: Natural Language Navigation and Spatial Reasoning in Visual Street Environmentscited by 239
  17. 2022Measuring Progress on Scalable Oversight for Large Language Modelscited by 225
  18. 2017Learning with Opponent-Learning Awarenesscited by 222
  19. 2018A Survey and Critique of Multiagent Deep Reinforcement Learningcited by 177
  20. 2020Emergent Multi-Agent Communication in the Deep Learning Eracited by 170
  21. 2017Emergence of Language with Multi-agent Games: Learning to Communicate with Sequences of Symbolscited by 125
  22. 2018Gromov-Wasserstein Alignment of Word Embedding Spacescited by 122
  23. 2024When combinations of humans and AI are useful: A systematic review and meta-analysiscited by 106
  24. 2016Opponent Modeling in Deep Reinforcement Learningcited by 97
  25. 2024Functional Benchmarks for Robust Evaluation of Reasoning Performance, and the Reasoning Gapcited by 79
  26. 2020Multi-agent Communication meets Natural Language: Synergies between Functional and Structural Language Learningcited by 66
  27. 2018An Overview of Machine Teachingcited by 63
  28. 2018Talk the Walk: Navigating New York City through Grounded Dialoguecited by 44
  29. 2022A Survey of Ad Hoc Teamwork Researchcited by 43
  30. 2024Complementarity in Human-AI Collaboration: Concept, Sources, and Evidencecited by 35
  31. 2017Translating Neuralesecited by 16

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2024Position: Towards Bidirectional Human-AI Alignment
  2. 2026Align AI to Dynamic Human-AI Workflows
  3. 2025Human-AI Interaction Alignment: Designing, Evaluating, and Evolving Value-Centered AI For Reciprocal Human-AI Futures
  4. 2024Concept Alignment
  5. 2025Towards Integrated Alignment
  6. 2025Frictional Agent Alignment Framework: Slow Down and Don't Break Things
  7. 2024AI Alignment through Reinforcement Learning from Human Feedback? Contradictions and Limitations
  8. 2024On the Essence and Prospect: An Investigation of Alignment Approaches for Big Models