method
active
method:counterfactual-reflection-trainingCounterfactual reflection training
Fine-tuning on reflection turns to implant principles into workspace.
Neighborhood — ranked by edge-count
Papers (1)
paper
- Verbalizable Representations Form a Global Workspace in Language Modelsintroducesmentions
Methods (1)
method
- Counterfactual prediction trainingrelated_toTraining models to answer Yes/No whether a prompt edit changes their behavior.
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.
- Encoding possible future or alternative states; bioelectric patterns can represent an alternative morphological target even in an intact animal.
- The behavior that would have occurred had the value of a causal variable been different while everything else remained the same; used as training labels in DAS/MAS.
- The output value a model produces when an interchange intervention forces certain variables to take values from source inputs.
- Counterfactual reflection training's improvement is mediated by implanted J-space content.claim0.789Ablation reverts effect.
- The mental effort of holding models of how things could/should be different from actuality, contributing to compression stress.
- Ability to entertain competing hypotheses within one inference engine; proposed hallmark of mindful inference
- Lens for evaluating explanations by whether they help predict behavior on counterfactual inputs.