finding
active
finding:for-four-of-eight-tasks-arc-easy-arc-challenge-hellaswag-aqua-the-model-size-with-greatest-cot-accuracy-gain-equals-the-size-with-greatest-faithfulnessFor four of eight tasks (ARC Easy, ARC Challenge, HellaSwag, AQuA) the model size with greatest CoT accuracy gain equals the size with greatest faithfulness
Appendix C.
Source paper
extracted_from(2023) · Tamera Lanham · Anna Chen · Ansh Radhakrishnan · Benoit Steiner +26
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Explanation supported by §3.
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.
- Optimum at mid-size.
- Only task with degradation.
- Appendix C.
- Seven of eight tasks improve under CoT; mean CoT length is 4 steps with 89% between three and sixfinding0.785CoT statistics.
- Table 2.
- Table 2 early answering AOC.
- Lowest-AOC tasks.
- Task difficulty as the key variable distinguishing the two modes of CoT identified in the paper