Research map / Architectures and efficiency / Attention and transformers
On the Ability of Transformers to Verify Plans
Yash Sarrof, Yupei Du, Katharina Stein, Alexander Koller, Sylvie Thiebaux, Michael Hahn
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 |
|---|---|---|---|
| use well-formed variant instead of conditional-effects formulation of Lights Out (ablation) | GPT-2 plan verifier trained on plans of 11-100 actions · Lights Out (5x5), OOD plan length 101-200 | accuracy | Improved |
| use well-formed variant instead of standard STRIPS variant (ablation) | GPT-2 plan verifier · Colors, OOD plan length 101-200 | accuracy | Improved |
| use delete-free variant of the domain (ablation) | GPT-2 plan verifier · Heavy Grippers, OOD plan length and object count | accuracy | Improved |
| use well-formed variant of Heavy Grippers (ablation) | GPT-2 plan verifier · Heavy Grippers, more objects and longer plans | accuracy | Improved |
| increase number of objects at fixed plan length for STRIPS Colors | GPT-2 plan verifier · Colors STRIPS, plan length 100 | accuracy | Hurt |
Closest papers by meaning
- What Algorithms can Transformers Learn? A Study in Length Generalization (ICLR 2024 · similarity 0.89)
- Transformers Provably Learn Chain-of-Thought Reasoning with Length Generalization (NeurIPS 2025 · similarity 0.89)
- A Formal Framework for Understanding Length Generalization in Transformers (ICLR 2025 · similarity 0.88)
- Length Generalization for Transformers via Compression (NeurIPS 2026 · similarity 0.88)
- On the self-verification limitations of large language models on reasoning and planning tasks (ICLR 2025 · similarity 0.87)
- Universal Length Generalization with Turing Programs (ICML 2025 · similarity 0.87)
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.
