method
active
method:probe-based-data-attribution

Probe-Based Data Attribution

Linear classifier approach applied to model activations to identify which training datapoints caused undesired behaviors in post-training.

Neighborhood — ranked by edge-count

Thinkers (2)

thinker

Findings (5)

finding

Concepts (7)

concept
  • Activation space
    associated_with
    Representation space on which linear probes operate to attribute harmful behaviors to training data.
  • Post-Training
    implements
    The phase after pre-training where models are further tuned with techniques like DPO; the period where the studied behavior emerged.
  • Internal representations of the model on which probes operate; the method uses activations to rank datapoints.
  • Broader research area: methods to align model behavior after initial training, where undesired behaviors can emerge.
  • The task of attributing model behaviors to specific training datapoints.
  • Probes
    cites
    Interpretability tools that decode information from internal model activations; here, linear probes are used for data attribution.
  • Alignment approach that focuses on curating or modifying training data; the paper bridges this with interpretability methods.

Methods (1)

method
  • Simple linear classifiers trained on model activations used as the probing technique within the introduced method.

Datasets (1)

dataset
  • OLMo 2 7B
    implements
    Language model substrate on which probe-based data attribution was demonstrated and evaluated.

Artifacts (1)

artifact

Claims (1)

claim

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

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