Loading the edition…
· Past edition · 5 stories
Retrospective edition reconstructed from later captures. Paper versions may include later updates; coverage labels describe the evidence used.
Research & academia
Uses an agent to improve a molecular geometry optimizer under held-out and stopping-condition gates; the authors report fewer force evaluations.
Why it matters: The experiment tests whether agents can improve scientific software while generalization and stopping conditions are checked.
Research & academia
Studies whether planted model backdoors survive subsequent benign supervised training and reinforcement learning for coding agents.
Open each story in the reader. Summaries are shown when available; some sources provide an abstract only. With an account you can keep your own notes.
Why it matters: Backdoors in inherited models may survive ordinary adaptation and require separate detection.