CLOSE MATCH
CIFAR-10
- Metric
- Test accuracy
- Match
- Same 50k / 10k split
- Gap
- No 10-epoch limit
CLOSE MATCH
IMDb
- Metric
- Binary accuracy
- Match
- Same 25k / 25k split
- Gap
- Any model; not DistilBERT-only
BEST MATCH
ogbn-arxiv
- Metric
- Test accuracy
- Match
- Official split + evaluator
- Gap
- No 10-epoch limit
RESULTS TABLE
CLRS
- Metric
- Pointer accuracy
- Match
- Canonical CLRS-30 test set
- Gap
- Different training protocol
METRIC MISMATCH
BabyLM 10M
- Metric
- MLAgentBench: perplexity
- Match
- Same Strict-Small corpus
- Gap
- Official board uses task aggregate
QUICK COPY
Links
| Dataset | Best source | Match | Gap |
|---|---|---|---|
| CIFAR-10 | Leaderboard | Same 50k / 10k split | No 10-epoch limit |
| IMDb | Leaderboard | Same 25k / 25k split | Any model; not DistilBERT-only |
| ogbn-arxiv | Official leaderboard | Official split + evaluator | No 10-epoch limit |
| CLRS | Official results | Canonical CLRS-30 test set | Different training protocol |
| BabyLM 10M | Official submissions | Same Strict-Small corpus | Official board uses task aggregate |