Repository navigation
Adds independent forced-choice scoring (0.13.0) - #18
Merged
Merged
Conversation
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
Adds
ForcedChoiceModelConfig.scoring_mode="independent"for dual encoders. Each alternative receives a score from the same scalar head, and softmax converts those scores into choice probabilities. The existingjointmode remains the default; single encoding with independent scoring is rejected.Supports fixed, random-intercept, and participant-specific scoring heads, preserves unselected alternatives, validates option names, and persists the scoring mode in checkpoints. Fixes dual-encoder training, participant parameter optimization, and early-stopping restoration. Updates the gallery configuration interface and API documentation, and bumps the minor version to 0.13.0 with a changelog entry.
Motivation
Forced-choice models previously predicted logits from combined option representations. Independent scoring provides a shared acceptability function over individual alternatives, including the Bradley-Terry choice rule for two alternatives.
Type of Change
Checklist
Testing
-m 'not slow_model_training' --no-cov; two existing dependency warnings.git diff --checkpassed.