Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-08-03 · high
The same plans, moved into a different agent with zero code changes, lifted its score from 75.8% to 81.0%.
Why now: As of August 3, STAIR's transfer result adds evidence to the debate over whether coding-agent gains come from bigger models or better scaffolding.
STAIR turns old repair trajectories into reusable, multi-level plans and hits 81.2% Pass@1 on SWE-bench Verified with MiniMax M2.5, per the paper posted to arXiv.
That's ahead of the 79.2% GPT-5 gets on the same benchmark. Most repair agents treat every issue independently, discarding whatever procedural knowledge they picked up solving the last one.
The system builds a tree instead of starting fresh. Fine-grained diagnostic actions sit at the bottom, high-level strategies at the top, and it draws from both to write issue-specific prompts.
The transfer result matters most. Researchers took the same plans, made zero code changes, and dropped them into mini-SWE-agent v2, a structurally different agent. Its score jumped from 75.8% to 81.0%.
It's not automatic. Ablations show mixing abstraction levels beats sticking to any single level, and raw, unabstracted trajectories transfer far worse than the distilled plans do.
For builders running repair agents, the lesson is to keep old trajectories instead of discarding them. Abstract them into levels rather than feeding raw logs back in.
Each link below shares sources, entities, or timing with this story.
GPT competes with Claude / Shared entities / Same source domain / What happened next
Linked by a graph relationship (GPT competes with Claude); both cover GPT, MiniMax M2, Pass, SWE; reported by the same outlet (arxiv.org).
GPT competes with Claude / Shared entities / Earlier coverage
Linked by a graph relationship (GPT competes with Claude); both cover GPT, SWE, Verified; earlier GPT coverage from 2026-05-27.
GPT competes with Claude / Shared entity: GPT / Same source domain / Shared topic / Earlier coverage / Tension
Linked by a graph relationship (GPT competes with Claude); both cover GPT; reported by the same outlet (arxiv.org).
GPT competes with Claude / Shared entity: GPT / Same source domain / Shared topic / Earlier coverage
Linked by a graph relationship (GPT competes with Claude); both cover GPT; reported by the same outlet (arxiv.org).
GPT competes with Claude / Shared entities / Earlier coverage
Linked by a graph relationship (GPT competes with Claude); both cover GPT, SWE; earlier GPT coverage from 2026-07-08.
Linked by a graph relationship (GPT competes with Claude); both cover SWE, Verified; earlier SWE coverage from 2026-04-26.
Copilot uses GPT / Shared entity: SWE / Shared topic / Earlier coverage
Linked by a graph relationship (Copilot uses GPT); both cover SWE; overlapping topics (agent, plan).
Copilot uses GPT / Shared entity: GPT / Shared topic / Earlier coverage
Linked by a graph relationship (Copilot uses GPT); both cover GPT; overlapping topics (action, agent).