# GRP: Goal-Reversed Prompting for Zero-Shot Evaluation with LLMs

Source: [GRP: Goal-Reversed Prompting for Zero-Shot Evaluation with LLMs](<https://www.sota2.com/research/paper/68fa77601e209131cea72d17>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-14T05:00:13.174Z

Authors: Mingyang Song, Mao Zheng, Xuan Luo

Published: 2025-03-08T09:44:24.000Z

## Abstract

Pairwise LLM-as-a-judge evaluation asks the judge to identify the \\emph{better} of two candidate answers. We study a one-line modification that asks for the \\emph{worse} answer instead and recovers the preference by elimination, a procedure we call Goal-Reversed Prompting (GRP). GRP introduces no extra inference rounds, composes with any prompt template (direct, chain-of-thought, or Arena-Hard SOP), and leaves the rest of the evaluation pipeline untouched. Two observations motivate the reversal. Reverse reasoning is a recurring strategy in human problem solving, and modern instruction-tuned judges exhibit a positive-leaning bias that asking for the worse answer can counteract. On JudgeBench under a strict consistency protocol that counts a judgment as correct only when both response orderings agree with the gold preference, GRP improves all three closed-source judges we test across both response-pair sources. With GPT-4o-generated pairs, the Arena-Hard SOP baseline improves from 61.71\\% to 66.23\\% for GPT-4o (+4.52) and from 60.00\\% to 66.00\\% for Claude-3.5-Sonnet (+6.00), with the largest absolute gains on Reasoning and Mathematics. The lift persists when response pairs come from Claude-3.5-Sonnet and when the SOP scaffolding is stripped to a minimal direct-prompting template, suggesting that goal reversal acts on the underlying judging behavior rather than on a particular rubric. Stronger judges benefit more than weaker ones, suggesting that goal reversal exposes additional reasoning capacity rather than compensating for its absence.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [LLM Judging on JudgeBench (GPT-4o 1.0 Generated Responses)](<https://www.sota2.com/research/sota/llm-judging-on-judgebench-gpt-4o-1-0-generated-responses.md>): LLM Judging; JudgeBench response pairs generated by GPT-4o 1.0; Knowledge: 68.18.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2503.06139>)

[PDF](<https://arxiv.org/pdf/2503.06139.pdf>)