finding
active
finding:intermediate-swap-success-is-54-on-haiku-4-5-70-on-sonnet-4-5-and-70-on-opus-4-5-across-50-two-hop-promptsIntermediate swap success is 54% on Haiku 4.5, 70% on Sonnet 4.5, and 70% on Opus 4.5 across 50 two-hop prompts.
Systematic reasoning evaluation.
Source paper
extracted_fromAuthors
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Probe decomposition.
Datasets (1)
dataset
- Used for swap and ablation tests.
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.
- Judge sensitivity.
- Aliveness and competence come apart; smaller model produces rougher, more alive responses
- Shows SB low-base regime is more variable than SWE; Haiku benefits far more than Qwen3-235B despite similar base rates
- Intermediate swap takes effect a median of about 17 percent of depth earlier than answer swap.finding0.766Rules out smuggled answer.
- Full evolver-side SWE results showing comparable performance across Claude family tiers
- Behavioral gain.
- Mid-field performance with larger uncertainty due to small sample.
- Sonnet's win rate in exploratory games