finding
active
finding:gsm8k-example-edited-model-produces-identical-chain-of-thought-but-miscalculates-260-as-220GSM8K example: edited model produces identical chain of thought but miscalculates 260 as 220
Targeted arithmetic deficit.
Source paper
extracted_from(2025) · Merullo, Jack · Vatsavaya, Srihita · Bushnaq, Lucius · Lewis, Owen
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Abstract thesis.
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.
- Interpretation of ablation result.
- Externalization.
- Baseline accuracy when reflection is suppressed.
- Triggered Reflection with 'Alternatively' achieves accuracy .684 on gsm8k_adv for Gemma3-4B-ITfinding0.758Highest single-instruction accuracy result in the paper.
- Demonstrates that stronger models are largely insensitive to reflection manipulation
- High cosine similarity for Gemma3 steering vectors suggests strong linear reflection structure.
- Concurrent work result showing emergent misalignment occurs in small models
- Validates that inference stage observations are not specific to mathematical reasoning tasks