finding
active
finding:deepseek-sentence-resampling-multipliers-4-9-at-s-5-3-1-at-s-30-1-1-at-s-100DeepSeek sentence resampling multipliers: 4.9× at S=5, 3.1× at S=30, 1.1× at S=100.
Efficiency gain on reasoning model.
Source paper
extracted_from(2026) · Eric Bigelow · Amir Zur · Satchel Grant · Tal Haklay +6
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Discussion.
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.
- Noise scaling.
- Reasoning > non-reasoning.
- Shows reasoning-focused models benefit most from VS in dialogue simulation tasks
- DeepSeek-V3.1 shows broad fine-tuning sensitivity; outputs code on nearly all open-ended prompts under insecure fine-tuning
- Evidence that reasoning length does not track safety performance
- Central interpretation.
- Llama sample-efficiency multiplier at 4-token spacing: 5.0× at S=5, 3.3× at S=30, 1.0× at S=100.finding0.747Efficiency gain.
- One DS-v3.2 trace shows extreme self-escalation, suggestive of treating own bid as competitor.