| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| PubMed | DP-FT | Next Token Accuracy42.33 | 40 | 8d ago | |
| QMSum | DP-FT | Next Token Accuracy47 | 32 | 1mo ago | |
| WildChat | DP-FT | Next Token Accuracy51 | 32 | 1mo ago | |
| BBC | DP-FT | Next Token Accuracy40.05 | 32 | 1mo ago | |
| OpenWebText | Concepts | PPL18.68 | 30 | 17d ago | |
| C4 (held-out) | Concepts | Perplexity (PPL)21.5 | 30 | 17d ago | |
| C4 | Concepts | OOD Perplexity21.1 | 30 | 17d ago | |
| OpenWebText (held-out) | Concepts | ID PPL18.53 | 30 | 17d ago | |
| py150 | TravTrans | MRR67.2 | 16 | 1mo ago | |
| Wikitext2 | LoRA (QV4) | Perplexity7.64 | 12 | 1mo ago | |
| internal | TravTrans | MRR0.615 | 10 | 1mo ago | |
| Pre-training corpus (train) | LIME+1 | Token Accuracy66.4 | 9 | 1mo ago | |
| WildChat | DP-FT | BERT-Small Next Token Accuracy (eps=inf)28.78 | 5 | 1mo ago | |
| BBC | DP-FT | Accuracy (ϵ=∞)25.75 | 5 | 1mo ago | |
| OpenWebText (val) | Hermite activation | Perplexity18.8 | 5 | 1mo ago | |
| OpenWebText (train) | Hermite activation | PPL18.4 | 5 | 1mo ago | |
| QMSum | DP-FT | Acc (BERT-Small, Epsilon=Inf)32.82 | 4 | 1mo ago | |
| Amber 1.2T tokens | LLAMA2-7B | BPD4.28 | 4 | 1mo ago | |
| Observed Antibody Sequences (OAS) processed subset | Mistral 94M | Bits Per Dimension1.62 | 4 | 1mo ago | |
| OpenWebText | GPT2 (124M) | BPD7.61 | 4 | 1mo ago | |
| blocksworld-8b (test) | Stick-Breaking Transformer | Accuracy99.8 | 3 | 1mo ago | |
| blocksworld 8b (train) | Stick-Breaking Transformer | Accuracy100 | 3 | 1mo ago | |
| SQuAD (test) | Standard APS | Mean |C|847 | 1 | 1mo ago |