claim
active
claim:extra-test-time-compute-alone-is-not-responsible-for-cot-s-performance-boostExtra test-time compute alone is not responsible for CoT's performance boost.
From filler token results.
Source paper
extracted_from(2023) · Tamera Lanham · Anna Chen · Ansh Radhakrishnan · Benoit Steiner +26
Neighborhood — ranked by edge-count
Papers (1)
paper
- Measuring Faithfulness in Chain-of-Thought Reasoningextendsmentions
Findings (1)
finding
- Evidence against test-time compute hypothesis.
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Additional test-time computation from longer context is responsible for CoT's performance boost.hypothesis0.925Tested via filler tokens.
- From decoupling of AOC and accuracy gain.
- Information encoded in phrasing choices of CoT is responsible for the accuracy boost.hypothesis0.765Tested via paraphrasing.
- Subsets amenable to monitoring.
- Scope generalization results after LoRA+CoT fine-tuning
- Practical implication.
- From paraphrasing results.
- Contrasts with Baker et al.