Fetching from the wire…
01
02
03
04
05
06
07
08
Skills2026-07-30 · source-backed
ExplainBench found explanation quality diverges from SWE-bench Verified scores, with agents recurrently asserting patch correctness that didn't hold, which is exactly the failure that defeats review by making a bad patch read as vetted. A dedicated audit agent running additional tests improved trustworthiness across every agent evaluated. Cheap second pass, high leverage.
Each link below shares sources, entities, or timing with this story.
The most useful AI-productivity dataset I've seen came from a company with every incentive to measure it honestly, because they're 3,500 people trying to run on their own product. The Pragmatic Engineer's July 29 deep dive inside Anthropic reports code output per engineer up 2...
FailFast-RestartSmart (arXiv 2608.03222) trains a small monitor on observable trajectory prefixes only, no policy logits or hidden states, to predict that a repo-level agent run is heading for failure, then launches a fresh rollout with the interrupted diff offered as an optio...
The study extracted 130 clean atomic state transitions from 707 real issues in SWE-bench Lite and Verified. Plain RAG scored 0.57-0.59 answer accuracy; an LLM reranker didn't help and added latency, about 18 seconds against 2.1. A (subject, relation, object) supersession memor...
Agent Lightning v1.0 (arXiv 2608.17528) inverts the standard agentic RL architecture, and the inversion is the whole point. Normally the training engine owns the environment loop. It drives the agent, collects trajectories, computes rewards. Which means your training setup and...
RETRACE has a verifier infer what problem the patch appears to solve using only the patch and trajectory, then compares that inference against the real issue. Training-free, lifted Pass@1 by 7.0% and 3.6% on mini-SWE-agent over SWE-bench Verified. The information-hiding trick...
The standard criterion for LLM-generated bug reproduction tests, fails on buggy code and passes on the golden fix, turns out to be insufficient. Many F→P tests are "lax": they reproduce the symptom while still admitting plausible-but-wrong patches. Worse, co-generating the tes...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.