finding
active
finding:llama-3-1-8b-replication-shows-similar-qualitative-trends-for-geometry-event-boundaries-noise-and-in-context-graph-tasks-saes-not-fully-trainedLlama-3.1-8B replication shows similar qualitative trends for geometry, event boundaries, noise and in-context graph tasks (SAEs not fully trained).
App. E.
Source paper
extracted_from(2025) · Lubana, Ekdeep Singh · Rager, Can · Hindupur, Sai Sumedh R. · Costa, Valerie +12
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Overall conclusion.
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.
- Llama-3.3-70B exhibits internal consistency-checking mechanisms that operate during inferenceclaim0.842Central interpretive claim of the paper supported by causal ablation and activation evidence
- Key mechanistic finding showing task-agnostic reuse of arithmetic circuitry
- Shows behavioral pattern of self-correction is trainable in smaller models
- Establishes generalizability of the core difficulty-boundary finding across model families.
- Larger models linearly represent more general concepts including truth
- Core empirical result demonstrating that manifold steering produces on-target, behavior-aligned outputs.
- Replication across open-weight models supports scale-emergence finding
- The specific Fourier feature periods identified confirm base-10 rather than modular computation