A global workspace in language models
Anthropic reports a small set of internal Claude activation patterns, the J-space, that behave like a global workspace, and examines whether they can be monitored for misbehavior.
Research
Anthropic
Research on reliable, interpretable and steerable AI. anthropic.com
Focus All
Topic Interpretability + Control
Source Anthropic · 7
Date Any time
1 item
1 item
Persona vectors : Monitoring and controlling character traits in language models
Anthropic identifies persona vectors, neural activity directions controlling model character traits like sycophancy, and shows they can monitor and steer undesirable personality shifts.
1 item
Putting up Bumpers
Argues that even without solving alignment, layered safeguards (interpretability audits, behavioral red-teaming, post-deployment monitoring) can catch and correct misaligned early AGI systems before they cause severe harm.
1 item
Insights on Crosscoder Model Diffing
Anthropic interpretability note on crosscoder model diffing finds exclusive features are polysemantic and dense, proposes designated shared features with reduced sparsity penalty to yield more monosemantic model-difference features.
2 items
Using Dictionary Learning Features as Classifiers
Anthropic interpretability team tests whether dictionary-learning features can train classifiers that detect bioweapon-related prompts, finding feature-based linear classifiers competitive with raw activations and useful for spotting spurious correlations.
Using dictionary learning features as classifiers
Anthropic interpretability team reports preliminary work using dictionary learning features as classifiers to detect harmful content or behaviors in model activations, for safeguards.
1 item
Measuring faithfulness in Chain-of-Thought reasoning
Tests whether chain-of-thought reasoning faithfully reflects model decisions by intervening on CoTs. Finds faithfulness varies by task and declines on most tasks as models scale.