Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Explainability Via Causal Self-Talk
Nicholas Andrew Roy, Junkyung Kim, Neil Charles Rabinowitz
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train with Causal Self-Talk (any variant) instead of ordinary decoder | Ord-Dec · DaxDucks one-hot | causal faithfulness | Improved |
| re-route z back to internal state (RR-Dec) instead of ordinary decoder | Ord-Dec · DaxDucks one-hot | causal faithfulness | No clear effect |
| use CST instead of baseline decoders | Ord-Dec and RR-Dec · DaxDucks one-hot | correlational faithfulness | No clear effect |
| use CST-RL (online interventions with RL) instead of other CST variants | CST · DaxDucks language | causal faithfulness | Improved |
| use CST-MR (memory reversion) variant | CST · DaxDucks language | causal faithfulness | Hurt |
| use language messages instead of one-hot messages | CST · DaxDucks | faithfulness | Hurt |
| inject message z' about tag location at each trial | CST agent · DaxDucks with shuffled tags | return | Improved |
Closest papers by meaning
- Towards Trustworthy Autonomous Systems via Conversations and Explanations (AAAI 2024 · similarity 0.87)
- Teaching Transformers Causal Reasoning through Axiomatic Training (ICML 2025 · similarity 0.86)
- ADAM: An Embodied Causal Agent in Open-World Environments (ICLR 2025 · similarity 0.86)
- ExPO: Unlocking Hard Reasoning with Self-Explanation-Guided Reinforcement Learning (NeurIPS 2025 · similarity 0.85)
- Unveiling Causal Reasoning in Large Language Models: Reality or Mirage? (NeurIPS 2024 · similarity 0.85)
- From an LLM Swarm to a PDDL-empowered Hive: Planning Self-executed Instructions in a Multi-modal Jungle (ICLR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
