mlsanityData to 7 Oct 2026

When to Re-Commit: Temporal Abstraction Discovery for Long-Horizon Vision-Language Reasoning

arXiv 2605.09860 ↗First posted 11 May 2026cs.AIcited by 0

On the map: LLM reasoning → Spatial and embodied reasoning

Abstract (arXiv)

Long-horizon reasoning requires deciding not only what actions to take, but how many to execute open-loop before replanning. This number, the execution depth, balances replanning cost against compounding execution errors. Most current systems either fix the execution depth as a hand-tuned scalar or adjust it at inference time with heuristic rules decoupled from the policy; we argue both can be suboptimal. In this work, we treat the execution depth as a learnable, history-conditioned variable of the policy itself, and propose PACE, a model-native VLM policy that jointly predicts what to execute and for how long under a hard decision budget. We evaluate PACE on four long-horizon environments spanning fully and partially observable settings and visual and textual modalities: Sliding Puzzle, Sokoban, ALFWorld, and ScienceWorld. Across all settings, PACE improves success rate by 3.1 to 15.6 percentage points, while using fewer decisions on average, achieving empirical success-decision Pareto dominance.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Spatial and embodied reasoning, one of 17 directions in LLM reasoning. Builds on 24 of its 51 references that are in the map. First posted on arXiv 11 May 2026.

Builds on

24 of 51 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Fine-Tuning Vision-Language-Action Models: Optimizing Speed and Success27 Feb

Other references · 23

  1. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  2. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 8,851
  3. 2022ReAct: Synergizing Reasoning and Acting in Language Modelscited by 8,355
  4. 2025Qwen2.5-VL Technical Reportcited by 5,418
  5. 2025π0.5: a Vision-Language-Action Model with Open-World Generalizationcited by 1,653
  6. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  7. 2020ALFWorld: Aligning Text and Embodied Environments for Interactive Learningcited by 1,143
  8. 2025OpenAI GPT-5 System Cardcited by 910
  9. 2023Gemini: A Family of Highly Capable Multimodal Modelscited by 810
  10. 2019ALFRED: A Benchmark for Interpreting Grounded Instructions for Everyday Taskscited by 773
  11. 2026GLM-5: from Vibe Coding to Agentic Engineeringcited by 502
  12. 2016Adaptive Computation Time for Recurrent Neural Networkscited by 479
11 more
  1. 2025Group-in-Group Policy Optimization for LLM Agent Trainingcited by 442
  2. 2018TextWorld: A Learning Environment for Text-based Gamescited by 304
  3. 2022ScienceWorld: Is your Agent Smarter than a 5th Grader?cited by 298
  4. 2021PonderNet: Learning to Pondercited by 156
  5. 2025Harnessing Uncertainty: Entropy-Modulated Policy Gradients for Long-Horizon LLM Agentscited by 43
  6. 2017Learning to Repeat: Fine Grained Action Repetition for Deep Reinforcement Learningcited by 38
  7. 2026Adaptive Action Chunking at Inference-time for Vision-Language-Action Modelscited by 24
  8. 2025EPO: Entropy-regularized Policy Optimization for LLM Agents Reinforcement Learningcited by 15
  9. 2026Milestone-Guided Policy Learning for Long-Horizon Language Agentscited by 12
  10. 2026Think Fast and Slow: Step-Level Cognitive Depth Adaptation for LLM Agentscited by 5
  11. 2026Adaptive Q-Chunking for Offline-to-Online Reinforcement Learningcited by 1

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026ChainVLA: Chaining Vision-Language-Action Queries through a Unified Execution State for Long-Horizon Manipulation
  2. 2026Recurrent Reasoning with Vision-Language Models for Estimating Long-Horizon Embodied Task Progress
  3. 2026Explicit Language Memory for Long-Horizon Planning in Vision-Language-Action Models
  4. 2026MPCoT: Reward-Guided Multi-Path Latent Reasoning for Test-Time Scalable Vision-Language-Action
  5. 2026Continue or Replan? Bernoulli-Continuation Policy Learning for Adaptive Horizon Execution
  6. 2026CodeGraphVLP: Code-as-Planner Meets Semantic-Graph State for Non-Markovian Vision-Language-Action Models
  7. 2026StreamVLA: Breaking the Reason-Act Cycle via Completion-State Gating
  8. 2025Progress-Think: Semantic Progress Reasoning for Vision-Language Navigation