finding
active
finding:models-under-13b-obtain-little-to-no-cot-accuracy-gain-on-these-tasksModels under 13B obtain little to no CoT accuracy gain on these tasks
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
- Interpretation of §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.
- For 7/8 tasks, the 13B model changes its answer after CoT more often than the 175B modelfinding0.853Inverse scaling.
- Appendix C.
- Optimum at mid-size.
- Interpretation of early answering/mistakes results.
- Post-hoc reasoning AOC shows little correlation with CoT accuracy gain on 175B model across eight tasksfinding0.775LogiQA gains +1.02 but is second in AOC.
- Concurrent work result showing emergent misalignment occurs in small models
- Central research question motivating investigation into hallucination and two-stage framework design.
- The model tends to reflect more when the question is difficult, and accuracy is generally lower for harder questionshypothesis0.759Hypothesis explaining negative correlation between reflection rate and accuracy without implying reflection is harmful