Research map / Vision-language and multimodal / Robot action and navigation
PALM: Progress-Aware Policy Learning via Affordance Reasoning for Long-Horizon Robotic Manipulation
Yuanzhe Liu, Jingyuan Zhu, Yuchen Mo, Gen Li, Xu Cao, Jin Jin, Yifan Shen, Zhengyuan Li, Tianjiao Yu, Wenzhen Yuan, Fangqiang Ding, Ismini Lourentzou
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 |
|---|---|---|---|
| add progress prediction (ablation) | PALM · CALVIN ABC->D | average length: 4.02 -> 4.48 | Improved |
| add affordance foresight prediction (ablation) | PALM (fine-tuning) · CALVIN ABC->D | average length: 3.58 -> 4.48 | Improved |
| add progress prediction during pre-training (ablation) | PALM (pre-training) · CALVIN ABC->D | average length: 3.73 -> 4.48 | Improved |
| add inverse dynamics prediction (ablation) | PALM · CALVIN ABC->D | average length: 3.92 -> 4.48 (fine-tuning); 4.17 -> 4.48 (pre-training) | Improved |
| add Global affordance to vanilla VLA (ablation) | vanilla VLA · LIBERO-Long | success rate: 77.0 -> 84.0 | Improved |
| add Local affordance on top of Global (ablation) | VLA + Global · CALVIN ABC->D / LIBERO-Long | average length / success rate: CALVIN 3.96 -> 4.16; LIBERO-Long 84.0 -> 82.5 | Mixed |
| add Spatial affordance on top of Global and Local (ablation) | VLA + Global + Local · LIBERO-Long | success rate: 82.5 -> 86.5 | Improved |
| add Dynamic affordance to complete set (ablation) | VLA + G,L,S · LIBERO-Long | success rate: 86.5 -> 91.8 | Improved |
| include in-the-wild robot data (DROID, BridgeData V2) in pre-training (ablation) | PALM · LIBERO-Long | success rate: 73.5 -> 91.8 | Improved |
| include long-horizon video data in pre-training (ablation) | PALM · CALVIN ABC->D | average length: 3.73 -> 4.48 | Improved |
| include human-annotated fine-tuning data (ablation) | PALM · CALVIN ABC->D | average length: 3.58 -> 4.48 | Improved |
| include simulation benchmark pre-training data (ablation) | PALM · CALVIN ABC->D | average length: 3.96 -> 4.48 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- CoT-VLA — CoT-VLA: Visual Chain-of-Thought Reasoning for Vision-Language-Action Models (CVPR 2025) · better on 2 of 2 shared results
- Seer — SEER: Backdoor Detection for Vision-Language Models through Searching Target Text and Image Trigger Jointly (AAAI 2024) · better on 2 of 2 shared results
- CoA-VLA — CoA-VLA: Improving Vision-Language-Action Models via Visual-Text Chain-of-Affordance (ICCV 2025) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Affordance Field Intervention: Enabling VLAs to Escape Memory Traps in Robotic Manipulation (CVPR 2026 · similarity 0.92)
- CoA-VLA: Improving Vision-Language-Action Models via Visual-Text Chain-of-Affordance (ICCV 2025 · similarity 0.91)
- Affordance-R1: Reinforcement Learning for Generalizable Affordance Reasoning in Multimodal Large Language Models (AAAI 2026 · similarity 0.90)
- RoboCerebra: A Large-scale Benchmark for Long-horizon Robotic Manipulation Evaluation (NeurIPS 2025 · similarity 0.89)
- AffordDP: Generalizable Diffusion Policy with Transferable Affordance (CVPR 2025 · similarity 0.89)
- RoboProcessBench: Benchmarking Process-Aware Understanding in Visual-Language Robotic Manipulation (NeurIPS 2026 · similarity 0.89)
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.
