Convergence-Time Basin Mapping Probe

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• Probe-Based Data Attributionmethod0.723
  Linear classifier approach applied to model activations to identify which training datapoints caused undesired behaviors in post-training.
• Fixed Point Convergenceconcept0.718
  The behavior where repeated application of a transformer block converges to a state where X' = S_k(X')
• Probesconcept0.712
  Interpretability tools that decode information from internal model activations; here, linear probes are used for data attribution.
• Probe-based data attribution for alignmentconcept0.711
• The model converges to a more stable truth direction in middle-to-late layers, as evidenced by increasing cosine similarity between layer-wise probes.claim0.709
  Supported by the geometric transition visible in cosine similarity heatmaps for F0-F3.
• Probe Generalizationconcept0.708
  The ability of probes trained on one dataset to transfer accurately to topically and structurally different datasets
• Representational Convergenceconcept0.706
  The central empirical phenomenon: different neural networks trained on different data/objectives develop increasingly similar representations
• Probe-based method bridges interpretability (probes/activations) with data-centric alignment workclaim0.700
  Assertion from the paper's notes that the work connects two previously separate areas: interpretability tools and data-centric alignment.