distilbert·released 2022-03-021 source
distilgpt2 benchmark scores: 1 benchmark tracked, leading 1. 0% of its scores are independently reproduced — the rest are self-reported or unverified.
Leads on 1 of 1 benchmarks·best result 21.1 on WikiText-103 (unverified)·0 of 1 independently reproduced
Leads on: WikiText-103
| Benchmark | Score | Evidence status | Measured |
|---|---|---|---|
| WikiText-103 | 21.1perplexity | unverifiedT2 | — |
Claims drawn from cited facts, not live model generation.
This model, released in 2022 by distilbert, is a compact 0.09-billion-parameter model, placing it at the small end where the trade-off is efficiency and lower cost rather than maximum headroom.
3 cited facts
This model is scored on a single tracked benchmark, and it currently holds the top score there. Its record remains unverified: no outside evaluation has either confirmed or disputed the reported results, so the numbers should be read as neither validated nor contested.
3 cited facts
On average it trails the SOTA leader by 0.0 points under the disclosed harness, a score gap that reads as frontier-level (roughly a couple of points or less) and is not a capability claim. It holds the top score on 1 benchmark, which is genuine category leadership on that benchmark.
2 cited facts
2 facts cross-checked across data sources: 2 single-source.
Source facts, citations, and refresh stamp for this record.
Sources: huggingface_models
distilgpt2 is an AI model developed by distilbert, released 2022-03-02. Its benchmark record below tags every score as independently reproduced, self-reported, or unverified.
distilgpt2 has recorded scores on 1 benchmark, each shown with its evidence status.
distilgpt2 has recorded scores on 1 benchmark — WikiText-103. The full table above shows each score with its evidence status.
0 of 1 recorded scores (0%) are independently reproduced rather than self-reported by the lab.
distilgpt2 has 1 tracked claim: 1 unverified.