# Two-Stage Prompt Optimization for Few-Shot Relation Extraction: From Reasoning-Guided Search to Gradient-Guided Refinement

arXiv: https://arxiv.org/abs/2606.29639 · HTML: https://mlsanity.com/paper/2606.29639/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning.
- Builds on 17 of its 28 references that are in the map.
- First posted on arXiv 28 Jun 2026.

## Placed in

- LLM reasoning / [Prompting and chain-of-thought reasoning](https://mlsanity.com/lr/prompting-and-chain-of-thought-reasoning/)

## Builds on (most cited first, 17 of 17)

- [Quantifying Language Models' Sensitivity to Spurious Features in Prompt Design or: How I learned to start worrying about prompt formatting](https://mlsanity.com/paper/2310.11324/) (2023)
- [Automatic Prompt Optimization with "Gradient Descent" and Beam Search](https://mlsanity.com/paper/2305.03495/) (2023)
- [Prompt-aligned Gradient for Prompt Tuning](https://mlsanity.com/paper/2205.14865/) (2022)
- [Hard Prompts Made Easy: Gradient-Based Discrete Optimization for Prompt Tuning and Discovery](https://mlsanity.com/paper/2302.03668/) (2023)
- [FewRel: A Large-Scale Supervised Few-Shot Relation Classification Dataset with State-of-the-Art Evaluation](https://mlsanity.com/paper/1810.10147/) (2018)
- [PromptAgent: Strategic Planning with Language Models Enables Expert-level Prompt Optimization](https://mlsanity.com/paper/2310.16427/) (2023)
- [EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizers](https://mlsanity.com/paper/2309.08532/) (2023)
- [KnowPrompt: Knowledge-aware Prompt-tuning with Synergistic Optimization for Relation Extraction](https://mlsanity.com/paper/2104.07650/) (2021)
- [How to Unleash the Power of Large Language Models for Few-shot Relation Extraction?](https://mlsanity.com/paper/2305.01555/) (2023)
- [Revisiting Few-shot Relation Classification: Evaluation Data and Classification Schemes](https://mlsanity.com/paper/2104.08481/) (2021)
- [C-ICL: Contrastive In-context Learning for Information Extraction](https://mlsanity.com/paper/2402.11254/) (2024)
- [Chain of Thought with Explicit Evidence Reasoning for Few-shot Relation Extraction](https://mlsanity.com/paper/2311.05922/) (2023)
- [GReaTer: Gradients over Reasoning Makes Smaller Language Models Strong Prompt Optimizers](https://mlsanity.com/paper/2412.09722/) (2024)
- [Best of Both Worlds: A Pliable and Generalizable Neuro-Symbolic Approach for Relation Classification](https://mlsanity.com/paper/2403.03305/) (2024)
- [Automatic Prompt Optimization for Knowledge Graph Construction: Insights from an Empirical Study](https://mlsanity.com/paper/2506.19773/) (2025)
- [Local Prompt Optimization](https://mlsanity.com/paper/2504.20355/) (2025)
- [Structured Semantic Information Helps Retrieve Better Examples for In-Context Learning Applied to Few-Shot Relation Extraction](https://mlsanity.com/paper/2601.20803/) (2026)

## Abstract (arXiv)

Automatic prompt optimization is still underexplored for episodic few-shot relation extraction with smaller language models. We propose a two-stage framework that combines reasoning-based prompt optimization with gradient-based prompt optimization. The first stage can use any reasoning-based optimizer to make broadprompt improvements in natural language. The second stage applies our GradPO, which uses loss and gradient signals to identify high-impact prompt spans and refine them with local edits. Experiments on FS-TACRED and FS-FewRel show that local refinement usually improves prompts found by the first stage, and GradPO is the most consistent refiner. Our framework achieves state-of-the-art performance on FS-TACRED with Qwen3-4B and remains competitive on FS-FewRel.