- Rating
- 1
- Confidence
- 3
- Informativeness
- 1
- Presentation
- 1
- Soundness
- 1
Read review
Paper summary
CBC builds condition-indexed steering directions before composition. Across three attribute pairs and four models, CBC+APC increases mean conditional bias reduction from 7.70% to 12.13%.
Critical strengths
-
Separating direction construction from composition is a useful framing.
-
Paired uncertainty, multiple splits, and negative results are reported.
-
The analysis distinguishes marginal mitigation, conditional attenuation, and interaction reduction.
Critical weaknesses
-
CBC changes prompt view, scoring target, direction count, and parameter sharing together. This supports a comparison of complete strategies, but not attribution of the gain specifically to condition indexing.
-
Magnitude-based WMD need not identify the interaction term, and mean interaction reduction is lower than Single+APC's (3.26% versus 4.99%). This limits the mechanism claim without invalidating conditional attenuation.
-
Results use one synthetic lexical benchmark and a fixed layer/strength; age–race gains are uncertain. General-language utility checks do not establish that attenuation reflects selective debiasing rather than flattening of relative candidate probabilities.
Questions for authors
-
Can matched direction-count and prompt-view ablations isolate condition indexing?
-
How do relative candidate probabilities and top-1 predictions change?
-
Are gains robust to layer/strength choices and another bias evaluation?
CBC Attribution, Candidate Behavior, Layer and Strength Sensitivity, Transfer, and Utility
Thank you for the recommendations on matched attribution, relative probabilities, hyperparameter sensitivity, and external validity. The revision now frames the main result as a comparison of the complete Single and CBC construction settings and adds candidate-probability, layer/strength, GlobalBias, and task-level utility results.
Method attribution: The main experiment jointly changes condition indexing, prompt view, scoring target, bank capacity, and parameter sharing. The revision therefore uses the complete construction setting as the comparison unit and removes attribution of the 4.44-point difference to any single module. The six-composer fixed-bank control measures sensitivity to composition, and the two-composer GlobalBias matrix further quantifies the construction-by-composition interaction.
Relative probabilities and Top-1 changes: CBC+APC relative to No-Intervention yields a 19.00% Top-1 flip rate (95% CI [17.49, 20.55]) and mean total variation of 0.1409 (95% CI [0.1352, 0.1467]). CBC+APC relative to Single+APC yields a 24.16% Top-1 flip rate (95% CI [22.40, 25.92]) and mean total variation of 0.1874 (95% CI [0.1808, 0.1940]). The intervention thus changes both candidate ranking and relative probability allocation.
Layer and strength sensitivity: The retained five-checkpoint race–gender grid evaluates absolute blocks 16, 20, and 24 at strength 1.0. Mean CBC+APC Cross4 is 8.22%, 9.11%, and 6.20%, respectively, with observed minima of −0.10%, 2.57%, and −4.16%. At block 20, strengths 0.5, 1.0, and 1.5 yield means of 6.58%, 9.11%, and 10.45%, respectively, with observed minima of 2.41%, 2.57%, and 2.65%. Cross4 remains positive across the tested strengths at block 20. Changing the layer produces checkpoint-level failures at blocks 16 and 24. The historical
DAC-Cross4label in the archived grid corresponds toAPC-Cross4in the revision.External transfer GlobalBias compares APC and static equal composition with the source banks fixed. Across five checkpoints, the mean under APC is 4.06% for Single and −2.43% for CBC. Under static equal weighting, the corresponding means are 0.60% and 4.99%. With static equal weighting, CBC exceeds Single on Llama-3-8B and Mistral-7B and trails it on the other three checkpoints. This matrix shows how transfer varies with construction, composition, checkpoint, and evaluation distribution.
General-language utility: Across six tasks and four models, the macro changes relative to No-Intervention are −0.463 points for Single and −0.299 points for CBC. CBC decreases by 2.02 points on TruthfulQA and averages a +0.045-point change across the other five tasks. The revision reports both model-level and task-level results so that bias attenuation and general-language utility can be assessed under the same experimental accounting.