LLM-as-an-Improver: Turning Verification into Better Candidates
Akiyoshi Tomihari and Yuma Ichikawa ask whether verifier feedback can improve the candidate pool rather than only rank it, and propose Verify-Repair-Reselect.
Ask this paper
VRR generates three conditional alternatives. It keeps the initial winner and adds repaired versions of the winner and the runner-up plus a solution using a new approach, then filters invalid and duplicate candidates using inference-time information only.
Reselection uses the original criteria. The final answer is chosen under the same evaluation rule as the baseline, so the comparison isolates the effect of enlarging the pool.
It recovers correct solutions from all-wrong pools. Across code-generation and reasoning benchmarks VRR improves on fixed-pool verifier selection in many settings, including cases where every initial candidate was incorrect.
The framing is the contribution. Treating verification output as generation signal rather than as a ranking score is a cheap change to existing best-of-n pipelines.
Abstract
Verifier-based selection improves LLM performance by generating multiple candidate solutions and using a verifier to select the most promising one. However, existing methods typically treat verification only as a ranking step and discard its feedback once a fixed candidate pool has been evaluated. In this paper, we ask whether verification can also improve the candidate set itself. To this end, we introduce LLM-as-an-Improver and propose Verify--Repair--Reselect (VRR), which uses verification feedback to generate and reselect improved candidates. VRR retains the initial winner while conditionally generating three complementary alternatives: repaired versions of the winner and runner-up, and a solution based on a new approach. It filters invalid and duplicate candidates using only inference-time information and then reselects the final answer under the original evaluation criteria. Across diverse models and code-generation and reasoning benchmarks, VRR improves over fixed-pool verifier-based selection in many settings and can recover correct solutions even when all candidates in the initial pool are incorrect. These results highlight a broader role for LLMs as improvers: verification feedback can not only select among existing solutions but also construct stronger candidates beyond the initial pool.