# What Should the Reflector See? An Empirical Study of Evidence in Reflective Prompt Optimization

arXiv: https://arxiv.org/abs/2609.32452 · HTML: https://mlsanity.com/paper/2609.32452/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 97 references that are in the map.
- First posted on arXiv 26 Sep 2026.

## Builds on (most cited first, 24 of 24)

- [Language Models are Few-Shot Learners](https://mlsanity.com/paper/2005.14165/) (2020)
- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [Large Language Models are Zero-Shot Reasoners](https://mlsanity.com/paper/2205.11916/) (2022)
- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering](https://mlsanity.com/paper/1809.09600/) (2018)
- [Self-Refine: Iterative Refinement with Self-Feedback](https://mlsanity.com/paper/2303.17651/) (2023)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problems](https://mlsanity.com/paper/2402.14008/) (2024)
- [Large Language Models Are Human-Level Prompt Engineers](https://mlsanity.com/paper/2211.01910/) (2022)
- [AlphaEvolve: A coding agent for scientific and algorithmic discovery](https://mlsanity.com/paper/2506.13131/) (2025)
- [Large Language Models as Optimizers](https://mlsanity.com/paper/2309.03409/) (2023)
- [DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines](https://mlsanity.com/paper/2310.03714/) (2023)
- [Automatic Prompt Optimization with "Gradient Descent" and Beam Search](https://mlsanity.com/paper/2305.03495/) (2023)
- [GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning](https://mlsanity.com/paper/2507.19457/) (2025)
- [Optimizing Instructions and Demonstrations for Multi-Stage Language Model Programs](https://mlsanity.com/paper/2406.11695/) (2024)
- [TextGrad: Automatic "Differentiation" via Text](https://mlsanity.com/paper/2406.07496/) (2024)
- [LiveBench: A Challenging, Contamination-Limited LLM Benchmark](https://mlsanity.com/paper/2406.19314/) (2024)
- [Generalizing Verifiable Instruction Following](https://mlsanity.com/paper/2507.02833/) (2025)
- [StraGo: Harnessing Strategic Guidance for Prompt Optimization](https://mlsanity.com/paper/2410.08601/) (2024)
- [MASPOB: Bandit-Based Prompt Optimization for Multi-Agent Systems with Graph Neural Networks](https://mlsanity.com/paper/2603.02630/) (2026)
- [C-Evolve: Consensus-based Evolution for Prompt Groups](https://mlsanity.com/paper/2509.23331/) (2025)

## Abstract (arXiv)

Reflective prompt optimization revises instructions using examples of a model's behavior, but which evidence the reflector should receive remains unclear. We study evidence composition, visibility of examples, candidate selection and domain-knowledge policy within a single-parent Pareto-guided search. Using Qwen3.5-9B as both task model and reflector, we evaluate nine reflection strategies on five datasets. From these experiments, we find three distinct patterns. For performance improvement, Failures-only produces the largest mean test gain (+8.0 percentage points), while Balanced-mix and No-examples+Val share the best mean performance rank. For reflection effectiveness, No-examples achieves the best rank for improving sampled parents, yet yields only a 1.4-point mean test gain: local reflection success does not necessarily produce a stronger final prompt. For overfitting assessment, Failures-only and Balanced-mix share the lowest mean calibration-gap rank, while larger gaps on GPQA and IFBench show that calibration gains can overstate held-out improvement. This gap is a descriptive indicator, not a direct measure of overfitting. Together, these results show why reflection strategies should be assessed separately on final performance, parent improvement and calibration-to-test transfer.