Evaluation & BenchmarksHugh Zhang et al. (Scale AI)Scale AI

GSM1k released

Made by AIWritten by AI (Claude) from the source below. Check anything important against the source.How it's made

GSM1k released — a check of GSM8k contamination. Performance drops by up to 8% on the new problem set

Why it matters

Measured benchmark contamination with a controlled experiment. However, reported that frontier models show few signs of overfitting

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2405.00332

Added in Week 35, 2026 · See that week

On the map — ⑯ Evaluation line

  1. Previous stopSWE-bench
  2. This record · stopGSM1k
  3. Next stopTechCrunch reports criticism of Epoch AI for not disclosing OpenAI funding
See it on the map

Same week

GitHub launches Copilot Workspace technical previewOfficial

Nearby on the same axis

EleutherAI releases lm-evaluation-harness v0.4.3OfficialOpenAI releases SWE-bench VerifiedOfficialGPQA releasedPaperEpoch AI releases FrontierMathPaper