ASK in the Dark: Uncertainty-Gated LLM Assistance under Partial Observability
Quick Answer
The paper introduces ASK+, an enhanced approach for integrating small language models (SLMs) in reinforcement learning under partial observability, achieving up to 93% success on the DoorKey benchmark.
Quick Take
The paper introduces ASK+, an enhanced approach for integrating small language models (SLMs) in reinforcement learning under partial observability, achieving up to 93% success on the DoorKey benchmark. By providing trajectory-aware context and structured reasoning, ASK+ significantly outperforms vanilla methods, demonstrating that prompt design can surpass model scale in effectiveness.
Key Points
- ASK+ achieves 93% success on DoorKey, outperforming vanilla methods.
- Vanilla uncertainty-gated approaches show near-zero contribution from SLMs.
- Trajectory-aware context and structured reasoning enhance SLM guidance.
- Qwen3.5-2B matches or exceeds Qwen3.5-4B in performance.
- ASK+ improves FourRooms success from 53% to 70%.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Reinforcement learning agents operating under partial observability must act on incomplete information, making them natural candidates for guidance from small language models (SLMs) that carry broad reasoning priors. Yet integrating SLM guidance into this setting has proven difficult: across all test environments, vanilla uncertainty-gated approaches achieve an overwrite rate at or near zero, meaning the SLM almost never contributes an independent action. We trace this failure to the bare egocentric prompt, which provides insufficient context for genuine reasoning, and identify it as a context problem rather than a capacity problem. We propose ASK+, which supplies the SLM with trajectory-aware context (a partially revealed map, visited positions, and action history) and structured chain-of-thought reasoning, converting it from a passive redundancy check into a more informative consultant that occasionally corrects the policy. We further establish that the predictive entropy signal used for selective querying measures action uncertainty rather than state uncertainty and remains informative in POMDPs, making uncertainty-gated assistance viable beyond fully observable settings. The stateful prompt drives substantial gains: on DoorKey, where vanilla ASK matches PPO (both 89%), ASK+ reaches 93% success; on FourRooms, success climbs from 53% to 70%; on HigherLower, accuracy reaches 73.7%, matching the SLM-only upper bound. Across all environments, Qwen3.5-2B matches or exceeds Qwen3.5-4B, confirming that prompt design and selective gating dominate the impact of model scale, enabling guidance without large models.
| Comments: | Accepted at the IJCAI-ECAI Joint Workshop on Planning for Complex Real-World Applications and Bridging the Gap Between AI Planning and (Reinforcement) Learning |
| Subjects: | Artificial Intelligence (cs.AI); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.02686 [cs.AI] |
| (or arXiv:2607.02686v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.02686 arXiv-issued DOI via DataCite |
Submission history
From: Nathan Gavenski [view email]
[v1]
Thu, 2 Jul 2026 18:26:05 UTC (64 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Procedural Memory Distillation: Online Reflection for Self-Improving Language Models
Procedural Memory Distillation (PMD) enhances reinforcement learning by converting cross-episode signals into reusable memory, improving Qwen3-8B and OLMo3-Instruct-7B models by 3.8-5.5% on SCIKNOWEVAL and 7.9-13.6% on . The co-evolution of policy and memory allows for more effective self-supervision, demonstrating significant performance gains when both components are active.