Feature attribution (gradient-based) correlates 0.8 with ablation effects on the 'John' and 'Kobe' examples.

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

• Gradient-based data attributionmethod0.779
  Baseline method against which probe-based ranking is compared; more computationally expensive.
• The magnitude of the normalized gradients (choice of αk) plays an important role in performance.claim0.760
  Insight about gradient normalization scaling.
• Under reward shaping (G=100, H=-100, F=0), Active Inference scored 99.52, Bayesian RL 99.77, Q-learning 95.56, with nearly identical behavior between belief-based agents.finding0.757
  Table 2, row 3, showing equivalence when prior preferences match rewards.
• We hypothesize that degraded generalization on benchmarks like MMLU may reflect the computational demands of the tasks.hypothesis0.753
  Connecting the paper's task-difficulty findings to prior observations of weak generalization on complex QA benchmarks.
• Feature attribution via gradient dot product with SAE decodermethod0.753
  Computing attribution as the dot product of the output logit gradient with the SAE decoder weight, multiplied by feature activation.
• Attribution graph for 'the princess lost her crown' reveals a femaleness signal pathway from 'princess' through attentionfinding0.750
  One component of the minimal subnetwork for predicting 'her', discovered via VPD attribution graph.
• Attribution graph reveals a pathway that detects the verb 'lost' and upweights object pronounsfinding0.747
  Second component of the subnetwork for 'her', complementing the femaleness signal.
• Probe-based data attribution effectively reduces harmful behaviors via data interventionsclaim0.746
  Authors' central interpretive assertion that their method meaningfully mitigates unwanted behaviors.