finding
active
finding:multi-attempt-rate-peaks-at-2-7-around-0-3-below-threshold-in-boost-sweep-experimentMulti-attempt rate peaks at 2.7% around -0.3σ below threshold in boost sweep experiment
Quantitative characterization of ESR operating regime in boost level sweep
Source paper
extracted_from(2026) · Alex McKenzie · Keenan Pepper · Stijn Servaes · Martin Leitgab +5
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.
- Multi-attempt improvement rate peaks at 83% around -1.0σ below threshold in Llama-3.3-70Bfinding0.814Shows slightly weaker steering allows more successful corrections, characterizing optimal ESR conditions
- Supporting finding showing ESR is driven by both higher multi-attempt rates and comparable improvement rates
- Secondary metric: percentage of responses containing multiple attempts, separating surface from actual self-correction
- Shows behavioral pattern of self-correction is trainable in smaller models
- Best VS result in synthetic data generation for math, demonstrating downstream improvement through diversity
- Demonstrates ESR can be deliberately enhanced through prompting in the largest model
- Resampling at probe-selected spans yields 14.1% attempted hacks vs 1.5% control for GLM 5.2finding0.754Probe predictive in GLM.
- On list summation task, hacking 30% unsteered, 100% with desperate +0.05 or calm -0.05, 0% oppositefinding0.753Task-level effect.