Search results

Unnecessary re-evals, especially in single-task mode?

Status: Open.

#357 In gepa-ai/gepa;

· YonatanGideoni opened on May 15, 2026

Per-proposal reflection LLM cost tracking

Status: Open.

#331 In gepa-ai/gepa;

· lukeleeai opened on Apr 13, 2026

Design: Unified Reflection LM Pool — sessions, agents, and ComBEE as first-class reflection LMs

Status: Open.

#329 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Apr 11, 2026

Make iteration scheduling pluggable via IterationStrategy Protocol

Status: Open.

#324 In gepa-ai/gepa;

· sfc-gh-mishah opened on Apr 10, 2026

Cached eval still counts toward metric budget

Status: Open.

#316 In gepa-ai/gepa;

· Shangyint opened on Apr 6, 2026

Gepa speed very slow - black box example

Status: Open.

#296 In gepa-ai/gepa;

· Goulustis opened on Mar 28, 2026

docs: AIME example includes candidate prompt in side_info unnecessarily

Status: Open.

#289 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Mar 23, 2026

feat: auto-inject score into side_info when not already present

Status: Open.

#288 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Mar 23, 2026

feat: raise when every iteration fails consecutively (raise_on_exception=False)

Status: Open.

#281 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Mar 20, 2026

feat: prompt mutation rate — control how much GEPA rewrites the prompt each iteration

Status: Open.

#277 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Mar 19, 2026

feat: token budget awareness — track and enforce prompt size during optimization

Status: Open.

#272 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Mar 18, 2026

feat: semantic deduplication of accumulated prompt heuristics

Status: Open.

#271 In gepa-ai/gepa;

· LakshyAAAgrawal opened on Mar 18, 2026