Repository navigation
Refused tasks are not scored: rank on the tasks every model answered - #59
Merged
Merged
Conversation
A refusal is the provider's safety policy, not a protocol, so it says nothing about a model's ability to write one; counting it as 0 ranked models by their filters. Everywhere a score is shown: - headline: mean over the tasks every model answered (8 now): Opus 1.000, Fable 0.969, GPT-6.1 Sol 0.898, Sonnet 0.842 - beside it: each model's mean over all the tasks it answered, with n - refusals: their own column, "refused, not scored" The "refusals count as 0" mean is removed from the leaderboard (now ranks on the common tasks; second view: all tasks answered), the report generator and REPORT.md, Figure 7 (bars: common, all answered, easy, paper-only; cost against the headline), the README and the paper. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
A refusal is the provider's safety policy, not a protocol, so it can't count as a 0.
New scoring everywhere (leaderboard, REPORT.md, Figure 7, README, paper):
The "refusals count as 0" mean is gone from every surface. The README notes that each new model can only shrink the common set, which is why the all-answered mean is shown too.
🤖 Generated with Claude Code