finding
active
finding:tpp-and-ravel-prefer-higher-l0-400-unlike-scrTPP and RAVEL prefer higher L0 (>400), unlike SCR
Unexpected result.
Source paper
extracted_from(2025) · Adam Karvonen · Can Rager · Johnny Lin · Curt Tigges +11
Neighborhood — ranked by edge-count
Papers (1)
paper
Hypotheses (1)
hypothesis
- Explains TPP/RAVEL preferring high L0.
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.
- Task-dependent optimal sparsity.
- Cross-judge validation of the primary ESR finding across OpenAI, Alibaba, Anthropic, and Google judge models
- ReLU underperforms.
- Table 2.
- Quantitative vulnerability profile for Llama-3.1-8B showing AS dominance
- RLFR policy matches base on standard benchmarks (e.g., MMLU 67.2 vs 69.2, GSM8K 73.0 vs 72.3)finding0.719No capability degradation.
- Demonstrates that harness loading is necessary but not sufficient for harness benefit; cleanest separation of activation and adherence
- Vulnerability profile for Qwen3.5-27B showing near-zero AS vulnerability