Prompting Policies for Multi-step Reasoning and Tool-Use in Black-box LLMs with Iterative Distillation of Experience

• URL: http://arxiv.org/abs/2605.14443v1
• Date: Thu, 14 May 2026 06:38:19 GMT
• Title: Prompting Policies for Multi-step Reasoning and Tool-Use in Black-box LLMs with Iterative Distillation of Experience
• Abstract summary: We propose a framework for training learned prompting policies via iterative distillation of experience.<n>A lightweight prompter model is optimized to maximize task-specific rewards for a larger, frozen worker Large Language Models.<n>We demonstrate significant gains improving performance from 55% to 90% in logic-intensive reasoning and 74% to 91% in tool-use tasks.
• Score: 0.6047855579999899
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The shift toward interacting with frozen, "black-box" Large Language Models (LLMs) has transformed prompt engineering from a heuristic exercise into a critical optimization challenge. We propose a Reinforcement Learning (RL) framework for training learned prompting policies via iterative distillation of experience. In this architecture, a lightweight prompter model is optimized to maximize task-specific rewards for a larger, frozen worker LLM. By utilizing a contrastive experience buffer that couples scalar rewards with dense textual critiques, our approach effectively amortizes iterative prompt refinement into single-shot policy weights. Our experimental analysis focuses on the Big Bench Extra Hard (BBEH) and Tau-bench suites, covering a diverse range of multi-step reasoning and tool-use tasks. We demonstrate significant gains, improving performance from 55% to 90% in logic-intensive reasoning and 74% to 91% in tool-use tasks. Furthermore, we analyze the structural evolution of prompts, demonstrating how the policy discovers specialized algorithmic heuristics. We provide comprehensive comparisons against state-of-the-art evolutionary baselines like GEPA, showing that iterative distillation achieves superior performance with higher sample efficiency.

Related papers

• AIR: Adaptive Interleaved Reasoning with Code in MLLMs [26.910280225921934]
  interleaved reasoning with code to enhance multimodal language models (MLLMs) has become a pivotal research frontier.<n>This paper empowers MLLMs with adaptive interleaved reasoning capabilities through extended reinforcement learning training on code-augmented complex numerical tasks.<n>Experiments demonstrate that after Reinforcement Learning training with the group-constrained reward function, performance improves by an average of 6.1 percentage points (pp) on evaluation benchmarks.
  arXiv  Detail & Related papers  (2026-06-22T17:58:54Z)
• Retrieval-Augmented LLM Agents: Learning to Learn from Experience [16.248836438253814]
  We study how to train retrieval-augmented LLM agents to leverage retrieved trajectories in-context.<n>We establish a robust supervised fine-tuning (SFT) recipe using LoRA that outperforms several state-of-the-art agent training pipelines.<n>Our results demonstrate that this combined approach significantly improves generalization to unseen tasks.
  arXiv  Detail & Related papers  (2026-03-18T20:45:04Z)
• MM-HELIX: Boosting Multimodal Long-Chain Reflective Reasoning with Holistic Platform and Adaptive Hybrid Policy Optimization [103.74675519953898]
  Long-chain reflective reasoning is a prerequisite for solving complex real-world problems.<n>We build a benchmark consisting 1,260 samples of 42 challenging synthetic tasks.<n>We generate post-training data and explore learning paradigms for exploiting such data.
  arXiv  Detail & Related papers  (2025-10-09T17:53:58Z)
• Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training [121.5858973157225]
  We investigate the effects of prolonged reinforcement learning on a small language model across a diverse set of reasoning domains.<n>We introduce controlled KL regularization, clipping ratio, and periodic reference policy resets as critical components for unlocking long-term performance gains.<n>Our model achieves significant improvements over strong baselines, including +14.7% on math, +13.9% on coding, and +54.8% on logic puzzle tasks.
  arXiv  Detail & Related papers  (2025-07-16T17:59:24Z)
• Reasoning BO: Enhancing Bayesian Optimization with Long-Context Reasoning Power of LLMs [14.58429044330196]
  This paper designs Reasoning BO, a novel framework that leverages reasoning models to guide the sampling process in BO.<n> Reasoning BO provides real-time sampling recommendations along with critical insights grounded in plausible scientific theories.<n>We evaluate our approach across 10 diverse tasks encompassing synthetic mathematical functions and complex real-world applications.
  arXiv  Detail & Related papers  (2025-05-19T08:20:40Z)
• LLMQuoter: Enhancing RAG Capabilities Through Efficient Quote Extraction From Large Contexts [2.685668802278156]
  We introduce LLMQuoter, a lightweight, distillation-based model designed to enhance Retrieval Augmented Generation (RAG)<n>Built on the LLaMA-3B architecture and fine-tuned with Low-Rank Adaptation (LoRA) on a 15,000-sample subset of HotpotQA, LLMQuoter adopts a "quote-first-then-answer" strategy, efficiently identifying key quotes before passing curated snippets to reasoning models.<n>This workflow reduces cognitive overhead and outperforms full-context approaches like Retrieval-Augmented Fine-Tuning (RAFT), achieving over 20-point accuracy gains across both small and large language
  arXiv  Detail & Related papers  (2025-01-09T20:01:15Z)
• CoPS: Empowering LLM Agents with Provable Cross-Task Experience Sharing [70.25689961697523]
  We propose a generalizable algorithm that enhances sequential reasoning by cross-task experience sharing and selection. Our work bridges the gap between existing sequential reasoning paradigms and validates the effectiveness of leveraging cross-task experiences.
  arXiv  Detail & Related papers  (2024-10-22T03:59:53Z)
• QPO: Query-dependent Prompt Optimization via Multi-Loop Offline Reinforcement Learning [58.767866109043055]
  We introduce Query-dependent Prompt Optimization (QPO), which iteratively fine-tune a small pretrained language model to generate optimal prompts tailored to the input queries.<n>We derive insights from offline prompting demonstration data, which already exists in large quantities as a by-product of benchmarking diverse prompts on open-sourced tasks.<n> Experiments on various LLM scales and diverse NLP and math tasks demonstrate the efficacy and cost-efficiency of our method in both zero-shot and few-shot scenarios.
  arXiv  Detail & Related papers  (2024-08-20T03:06:48Z)
• Fine-Tuning Large Vision-Language Models as Decision-Making Agents via Reinforcement Learning [79.38140606606126]
  We propose an algorithmic framework that fine-tunes vision-language models (VLMs) with reinforcement learning (RL) Our framework provides a task description and then prompts the VLM to generate chain-of-thought (CoT) reasoning. We demonstrate that our proposed framework enhances the decision-making capabilities of VLM agents across various tasks.
  arXiv  Detail & Related papers  (2024-05-16T17:50:19Z)
• Monte Carlo Tree Search Boosts Reasoning via Iterative Preference Learning [55.96599486604344]
  We introduce an approach aimed at enhancing the reasoning capabilities of Large Language Models (LLMs) through an iterative preference learning process. We use Monte Carlo Tree Search (MCTS) to iteratively collect preference data, utilizing its look-ahead ability to break down instance-level rewards into more granular step-level signals. The proposed algorithm employs Direct Preference Optimization (DPO) to update the LLM policy using this newly generated step-level preference data.
  arXiv  Detail & Related papers  (2024-05-01T11:10:24Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.