thinker:paul-bogdanPaul Bogdan
Author; multi-token extension; cited for slot representations.
Authored papers (1)
A sparse, privileged subset of residual-stream directions in Claude models meets five functional criteria of a global workspace, and it is found by asking which representations the model is poised to verbalize. The Jacobian lens (J-lens) averages the first-order effect of each layer's activation on future token logits over 1,000 pretraining-like prompts, which corrects the logit lens's coordinate mismatch and yields the J-space, a sparse nonnegative subframe of about k≤25 token vectors. On Sonnet 4.5 the J-space carries at most 10% of activation variance, yet it governs verbal report, directed modulation, internal reasoning and flexible generalization. Swapping lens coordinates redirects two-hop answers in 54% (Haiku 4.5) to 70% (Sonnet 4.5, Opus 4.5) of trials. The J-space component of a concept probe flips answers in 61% of trials against 28% for the non-J remainder, and clamping the J-space drops that remainder to 6%. Ablating the top-10 J-lens vectors leaves MMLU, SQuAD and CoLA intact but degrades multi-hop reasoning, translation and sonnet writing to below unablated Haiku 4.5. The workspace occupies layers ~L38–92, and a 1% subset of attention heads broadcasts it. Applications follow: ablating eval-awareness tokens raises blackmail from 0/180 to 13/180 rollouts, and counterfactual reflection training cuts Haiku 4.5's fabrication dishonesty from 0.25 to 0.07 through J-space content. LLMs thus contain a readable, editable functional analogue of conscious access, with its limits set by single-token vocabulary and no recurrence.
More papers — OpenAlex / S2
Co-authors (12)
- Adam Pearce1 shared
- Anna Soligo1 shared
- Emmanuel Ameisen1 shared
- Isaac Kauvar1 shared
- Jack Lindsey1 shared
- Joshua Batson1 shared
- Mateusz Piotrowski1 shared
- Nicholas Sofroniew1 shared
- Rowan Wang1 shared
- Runjin Chen1 shared
- Subhash Kantamneni1 shared
- Wes Gurnee1 shared
Other inbound relations (1)
Recent mentions (1)
- papers-typedauthors-verbalizable-representations.md