Loading the edition…
· Past edition · 8 stories
Retrospective edition reconstructed from later captures. Paper versions may include later updates; coverage labels describe the evidence used.
Research & academia
A paper audits defects in a scientific-coding benchmark.
Why it matters: Benchmark repairs can change model scores, making evaluation quality as important as the reported ranking.
Research & academia
The authors study knowledge transfer, modality competition and early joint training, then validate selected findings with 13.5B mixture-of-experts models. The captured v2 abstract is a later revision from 2026-08-06.
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: Training design can change how visual and language objectives interact.
Research & academia
MirageBench evaluates user-profile over-inference across 12 models. The authors report a negative model-level correlation between self-assessed and judged errors, with a wide confidence interval.
Why it matters: Personalization needs evidence checks when models infer unsupported user attributes.