GLUE
Nine-task English NLU suite combining acceptability, sentiment, similarity, paraphrase, inference and coreference into one score.
RoBERTa reached 88.5 on the official nine-task leaderboard by 2019-07-25, above the conservative 87.1 non-expert human aggregate; SuperGLUE was introduced for more headroom.
Performance Timeline
Longitudinal progression of model scores against human baselines.Performance & Historical Trajectory
Empirical score progression across model release dates and evaluation rounds.
| Model | Score | Date | Source Type | Provenance |
|---|---|---|---|---|
| DeBERTa (TuringNLRv4) | 90.8% | 2021-01-06 | vendor-reported | Source ↗ |
| ELECTRA-Large | 89.4% | 2020-03-23 | vendor-reported | Source ↗ |
| T5 (ensemble) | 90.3% | 2019-10-23 | vendor-reported | Source ↗ |
| ALBERT (ensemble) | 89.4% | 2019-09-26 | vendor-reported | Source ↗ |
| RoBERTa ensemble | 88.5% | 2019-07-25 | vendor-reported | Source ↗ |
| XLNet (ensemble) | 88.4% | 2019-06-19 | vendor-reported | Source ↗ |
| MT-DNN | 82.7% | 2019-02-25 | vendor-reported | Source ↗ |
| BERT-Large | 80.5% | 2018-10-11 | vendor-reported | Source ↗ |
Human Baseline & Difficulty Horizon
Calibrated human reference points, specialist benchmarks, and ceiling thresholds.Estimated human performance across general language understanding evaluation tasks.
Metric & Scoring Methodology
Verification protocols, aggregation formulas, and specialized metric variants.GLUE Score (0-100)Dataset & Compute Cost
Evaluation volume, public availability, API pricing, and local hardware requirements.$5 – $20 USD for full benchmark evaluation run on frontier APIs.
1x NVIDIA RTX 4090 (24GB) or A100 (40GB/80GB) via vLLM / SGLang
How to Run & Reproduce
Standardized evaluation protocols, CLI commands, and reproducible runner templates.lm_eval --model hf --model_args pretrained=<model_path> --tasks glue --batch_size autoopencompass --datasets glue --models <model_config># Standard API Evaluation Loop
from openai import OpenAI
client = OpenAI()
response = client.chat.completions.create(
model="gpt-4o",
messages=[{"role": "user", "content": prompt}],
temperature=0.0,
)When publishing results for GLUE, always report the exact prompt template, few-shot exemplar ordering, sampling temperature (temperature=0), maximum reasoning budget tokens, and the precise timestamped model snapshot ID.
Contamination & Memorization Analysis
Audit of pretraining exposure risks, memorization vectors, and refresh policies.Static fixed snapshot
Public on web / HuggingFace