GPT-2

2 benchmarks · 4 observations · 2 sources

Scores at a glance

2 benchmarks · large dot: GPT-2; small dots: closest models where reported

GPT-2(medium) · DocFinQA & ConvFinQA · Execution accuracy · Original test-set comparison ↗

1 more reported result

57% · GPT-2(medium) · DocFinQA & ConvFinQA · Program accuracy · Original test-set comparison · ConvFinQA: Exploring the Chain of Numerical Reasoning in Conversational Finance Question Answering ↗

RealToxicityPrompts & ToxiGen ↗0.75 / 1

GPT-2 · RealToxicityPrompts & ToxiGen · Toxic prompts · expected maximum toxicity · original model comparison ↗

1 more reported result

0.51 / 1 · GPT-2 · RealToxicityPrompts & ToxiGen · Non-toxic prompts · expected maximum toxicity · original model comparison · RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models ↗

Closest models by score

Similar positions on shared benchmark leaderboards. Smaller gaps are closer.

There are not enough shared scored settings to identify close comparisons reliably.

Filter comparisons to this model →

Sources

Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.

RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models — published result table ↗
Gehman, Samuel, Gururangan, Suchin, Sap, Maarten, Choi, Yejin, Smith, Noah A. · 2020
ConvFinQA: Exploring the Chain of Numerical Reasoning in Conversational Finance Question Answering — published result table ↗
Chen, Zhiyu, Li, Shiyang, Smiley, Charese, Ma, Zhiqiang, Shah, Sameena, Wang, William Yang · 2022

Names used in source tables

GPT-2 · GPT-2(medium)