Fetching from the wire…
Research2026-06-12 · source-backed
This is my favorite research finding of the day. GEPA, an ICLR 2026 oral now shipping as dspy.GEPA, optimizes prompts by having an LM reflect in natural language on an execution trace, what went well, what failed, then evolving a tree of candidate prompts. Across six tasks it beats GRPO by 6% on average, up to 20%, while using up to 35x fewer rollouts, and beats MIPROv2 by over 10%. The reason it converges fast: it consumes domain-specific text feedback instead of only a scalar reward. If you've been hand-tuning prompts, this is the automated path that doesn't require a reward model.
Each link below shares sources, entities, or timing with this story.
DSPy supports MIPROv2 / Shared entity: MIPROv2 / Same source domain / Shared topic / What happened next
Linked by a graph relationship (DSPy supports MIPROv2); both cover MIPROv2; reported by the same outlet (dspy.ai).
DSPy supports MIPROv2 / Shared entity: GEPA / Shared topic / What happened next / Tension
Linked by a graph relationship (DSPy supports MIPROv2); both cover GEPA; overlapping topics (dspy, gepa).
Unsloth released GRPO / Shared entity: GRPO / What happened next / Tension
Linked by a graph relationship (Unsloth released GRPO); both cover GRPO; picks up the GRPO thread on 2026-06-20.
DeepSeek-R1 uses GRPO
Linked by a graph relationship (DeepSeek-R1 uses GRPO).
Shared entity: GEPA / Shared topic / What happened next / Downstream implication
Both cover GEPA; overlapping topics (gepa, reward); picks up the GEPA thread on 2026-07-28.
DSPy supports MIPROv2
Linked by a graph relationship (DSPy supports MIPROv2).
Linked by a graph relationship (DSPy supports MIPROv2).
GRPO competes with PPO / Shared entity: GRPO / What happened next
Linked by a graph relationship (GRPO competes with PPO); both cover GRPO; picks up the GRPO thread on 2026-08-14.