Issues

• Make the coding agent backend pluggable

  gepa-ai/gepa#459 · Filter by author ktalg opened on Sep 18, 2026
  2

• Bugs Identified Post 0.1.4

  gepa-ai/gepa#448 · Filter by author LakshyAAAgrawal opened on Sep 3, 2026
  2

• Refiner score is credited to the unrefined candidate

  gepa-ai/gepa#440 · Filter by author youph opened on Aug 26, 2026
  3

• Decoupling and parallelizing rollouts + reflections

  gepa-ai/gepa#438 · Filter by author maks-ivanov opened on Aug 24, 2026
  2

• Verify evals on Papers with Code

  gepa-ai/gepa#427 · Filter by author NielsRogge opened on Aug 14, 2026

• Project Submission

  gepa-ai/gepa#419 · Filter by author sh-patterson opened on Aug 3, 2026

• omni(meta-harness): Multi-objective search for meta-harness + Codex integration

  gepa-ai/gepa#417 · Filter by author rishiskhare opened on Aug 2, 2026

• Reference: GEPA public API surface & ecosystem usage (stability tiers for future refactors)

  gepa-ai/gepa#408 · Filter by author LakshyAAAgrawal opened on Jul 23, 2026

• ComBEE §3.3: dynamic batch size controller (adaptive reflection_minibatch_size)

  gepa-ai/gepa#395 · Filter by author LakshyAAAgrawal opened on Jul 16, 2026

• Add a run_dir admission manifest for candidate acceptance/rejection audits

  gepa-ai/gepa#381 · Filter by author sunghunkwag opened on Jun 17, 2026
  6

• Idea: Add Program Summarization Context to GEPA Instruction Proposer for Improved Module Role Alignment

  gepa-ai/gepa#380 · Filter by author LakshyAAAgrawal opened on Jun 17, 2026

• GEPA doesn't remember rejected proposals — re-sampling the same parent repeats the same failed mutation

  gepa-ai/gepa#379 · Filter by author LakshyAAAgrawal opened on Jun 17, 2026
  2

• Coding agents running GEPA via /goal don't know to set a high-enough max_metric_calls — proposed API + docs fixes

  gepa-ai/gepa#375 · Filter by author LakshyAAAgrawal opened on Jun 13, 2026
  1

• Question abour reflection_lm and task_lm

  gepa-ai/gepa#363 · Filter by author smileformylove opened on May 30, 2026

• Question about task_lm and reflection_lm

  gepa-ai/gepa#362 · Filter by author smileformylove opened on May 30, 2026

• Add first-class vLLM / Hugging Face model loading support

  gepa-ai/gepa#359 · Filter by author Milan933-coder opened on May 20, 2026
  1

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

  gepa-ai/gepa#357 · Filter by author YonatanGideoni opened on May 15, 2026
  1

• Per-proposal reflection LLM cost tracking

  gepa-ai/gepa#331 · Filter by author lukedhlee opened on Apr 13, 2026
  1

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

  gepa-ai/gepa#329 · Filter by author LakshyAAAgrawal opened on Apr 11, 2026
  2

• Make iteration scheduling pluggable via IterationStrategy Protocol

  gepa-ai/gepa#324 · Filter by author sfc-gh-mishah opened on Apr 10, 2026
  1

• Cached eval still counts toward metric budget

  gepa-ai/gepa#316 · Filter by author Shangyint opened on Apr 6, 2026
  2

• Gepa speed very slow - black box example

  gepa-ai/gepa#296 · Filter by author Goulustis opened on Mar 28, 2026

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

  gepa-ai/gepa#288 · Filter by author LakshyAAAgrawal opened on Mar 23, 2026

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

  gepa-ai/gepa#281 · Filter by author LakshyAAAgrawal opened on Mar 20, 2026

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

  gepa-ai/gepa#277 · Filter by author LakshyAAAgrawal opened on Mar 19, 2026