Verbalizable Representations Form a Global Workspace in Language Models Wes Gurnee Nicholas Sofroniew Jack Lindsey et al. (Anthropic)

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

 Verbalizable Representations Form a Global Workspace in Language Models 2026 Transformer Circuits Thread gurnee2026-global-workspace-llm gurnee2026-global-workspace-llm.txt, the project's staged copy f8862134eecf3870c1222d3331c02cd7981dd708 personal-copy artificial-intelligence

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

 0 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. system-or-tool-building Quotation withheld because this record is not cleared for redistribution. A CONSTRUCT IS DEFINED BY ANALOGY TO ANOTHER FIELD'S CRITERIA, AND THE ANALOGY IS THE DEFINITION RATHER THAN A MOTIVATION FOR IT: we Quotation withheld. What makes something a member of the new category is that it satisfies conditions imported from a theory about brains about this quote: The sentence begins with a word the extractor split - "W e" for "We" - so the span starts at the second word. This file has 469 such splits and no glyph codes at all, so nothing marks them but reading. PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. AND THE PAPER'S OWN LIMITATIONS SECTION SAYS THE DISTINCTION IT RELIES ON HAS NO PRINCIPLED DEFINITION: this judgment WAS SOMEWHAT POST-HOC, and WE DID NOT PROVIDE A PRINCIPLED DEFINITION OF WHAT DISTINGUISHES A “WORKSPACE” REPRESENTATION FROM A “motor” ONE. The boundary was drawn empirically and the criterion for it is declared absent, by the people who drew it, in the same paper that uses it throughout about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. THE MISSING CRITERION IS SPECIFIED AS A CHOICE BETWEEN TWO REPAIRS, NEITHER OF WHICH IS PERFORMED: Quotation withheld What would fix the vocabulary is stated exactly, and left as a description of a paper someone else could write about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. AND THE REASON THE CRITERION IS NEEDED IS GIVEN AS AN INABILITY, NOT A DIFFICULTY: the tasks were chosen BECAUSE THE FLEXIBLE/AUTOMATIC DISTINCTION IS CLEAR FOR THEM; however, for many tasks of practical interest **IT IS NOT**, and WE HAVE NO WAY OF PREDICTING IN ADVANCE, FOR AN ARBITRARY COMPUTATION, WHETHER IT WILL ENGAGE THE J-SPACE. The cases were selected for the property that the distinction applies to them, and that selection is disclosed about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. A CONCEPT IS DEFINED AS WHATEVER THE SYSTEM ITSELF LATER PRODUCES, AND THE CIRCULARITY IS THE POINT OF THE EXPERIMENT: Quotation withheld The definition has no content until the system has acted, which is what makes the test possible about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. A METHOD IS DEFINED BY THE QUESTION IT ASKS RATHER THAN BY WHAT IT MEASURES, AND CONTRASTED WITH THE METHOD IT REPLACES: the J-lens provides a complementary way of interpreting a feature - IT ASKS WHAT THE FEATURE DIRECTION IS VERBALIZABLE AS, INDEPENDENT OF WHERE IT FIRES. The older method is said to MISLEAD WHEN THE FEATURE ENCODES AN ABSTRACT ASSESSMENT WHOSE SURFACE TEXTUAL CORRELATES ARE INCIDENTAL about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. AN ANALOGY IS ASSERTED AND ITS LIMIT IS STATED IN THE SAME SENTENCE, WITH THE UNCLAIMED PART SPELLED OUT: Quotation withheld - specialized, encapsulated processors competing for entry to a workspace that broadcasts back to them through recurrent connections. The borrowed theory is named in full precisely where the borrowing stops about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. THE DISANALOGY IS PUT FIRST AND THE EXPECTATION DENIED: Quotation withheld The reason the finding would be surprising is given before the finding is defended about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. TWO EXPLANATIONS OF THE PAPER'S CENTRAL FINDING ARE OFFERED AND NEITHER IS CHOSEN: Quotation withheld The property named in the title is explained by the shape of the output channel rather than by anything about the representation about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. A SCOPE LIMIT IS STATED WITH A COINED METAPHOR THAT CONCEDES WHAT THE METHOD CANNOT SEE: Quotation withheld The instrument's reach is named as a part of its object, not as a defect of the study about this quote: PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. AND THE PROPERTIES THE ANALOGY DOES NOT ESTABLISH ARE LISTED AS ABSENT: Quotation withheld A borrowed framework is adopted in part, and the unadopted part is named rather than left for a reader to notice about this quote: The sentence begins with a word the extractor split - "W e" for "We" - so the span starts at the second word. This file has 469 such splits and no glyph codes at all, so nothing marks them but reading. PROVENANCE: this record belongs to the priming sub-corpus, which is the other terminal's lane, and these mechanisms were written by the project. No two hands were on the record - it carried no mechanisms before these, and the project confirmed on 2026-08-17 that the project is auditing rather than tagging. Kept rather than retracted on his instruction (17:22): retracting verified work over a bookkeeping flag would leave a retraction reason that misdescribes the row. Recorded here so a reader learns it from the corpus. This one was authorised in advance: the project named it at 16:39 - “bouchaud2008, gurnee2026, pelagios2023, dekel2021 - do those” - before the merge. The note is here because the lane fact is true regardless of who approved it. Quotation withheld because this record is not cleared for redistribution. what this limits: every workspace claim in the paper: THE JACOBIAN LENS IS AN IMPERFECT TOOL, WHICH WE BELIEVE ONLY APPROXIMATELY AND INCOMPLETELY CAPTURES THE MODEL'S UNDERLYING WORKSPACE STRUCTURE Quotation withheld because this record is not cleared for redistribution. what this limits: the completeness of any concept list the method produces: THE SET OF CONCEPTS THE LENS CAN NAME IS EXACTLY THE SET OF CONCEPTS THAT HAVE A SINGLE-TOKEN NAME IN THE TOKENIZER'S VOCABULARY - the instrument's lexicon bounds what is findable Quotation withheld because this record is not cleared for redistribution. what this limits: any structural or relational reading of the results: THE J-LENS TREATS THE WORKSPACE AS A FLAT COLLECTION OF INDEPENDENTLY ACTIVE CONCEPT VECTORS, so binding is not represented Quotation withheld because this record is not cleared for redistribution. what this limits: any readout taken from the lens: AT SOME POSITIONS AND LAYERS, THE TOP LENS TOKENS ARE ONES WE CANNOT MAKE SENSE OF, and the rate was not characterised systematically Quotation withheld because this record is not cleared for redistribution. what this limits: the early-layer result, where object and instrument cannot be separated: WE DO NOT KNOW WHETHER THIS IS A FACT ABOUT THE MODEL OR AN ARTIFACT OF THE LENS Quotation withheld because this record is not cleared for redistribution. what this limits: any mechanistic account of the workspace: WE HAVE NOT CHARACTERIZED WHAT CAUSES A REPRESENTATION TO ENTER IT Quotation withheld because this record is not cleared for redistribution. what this limits: any safety or alignment-monitoring use: WE DO NOT FEEL COMFORTABLE MAKING THE STRONGER CLAIM THAT MONITORING THE J-SPACE IS SUFFICIENT FOR ALIGNMENT MONITORING, OR THAT ANY SOPHISTICATED PLAN THE MODEL MIGHT EXECUTE MUST BE REPRESENTED THERE Quotation withheld because this record is not cleared for redistribution. what this limits: the generality across scale and training: WE HAVE STUDIED THE WORKSPACE PRIMARILY IN FULLY TRAINED PRODUCTION LLMs OF CONSIDERABLE SIZE Quotation withheld because this record is not cleared for redistribution. what this limits: the analogy to the neuroscience theory: Quotation withheld Quotation withheld because this record is not cleared for redistribution. eliminates: the possibility that the ablation did not do what it claims, so that a null downstream would be uninterpretable · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. eliminates: broad degradation of writing capability as the explanation of the drop in experiential language · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. eliminates: the effect arising from ablating any sufficiently large set of heads rather than these ones Quotation withheld because this record is not cleared for redistribution. eliminates: two alternatives at once - generic capability degradation from any perturbation, and regression to the mean from added noise · outcome (stated by the tagger, no quotation): held Quotation withheld because this record is not cleared for redistribution. global workspace Quotation withheld because this record is not cleared for redistribution. consciously accessible information Quotation withheld because this record is not cleared for redistribution. single-token name Quotation withheld because this record is not cleared for redistribution. J-space component Quotation withheld because this record is not cleared for redistribution. counterfactual reflection training Quotation withheld because this record is not cleared for redistribution. J-space Quotation withheld because this record is not cleared for redistribution. J-space component Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. a second reading DIFFERING from the recorded value. Recorded: existence — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is)., Quotation withheld because this record is not cleared for redistribution. a second reading DIFFERING from the recorded value. Recorded: system-or-tool-building — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is)., Quotation withheld because this record is not cleared for redistribution. a second reading DIFFERING from the recorded value. Recorded: global workspace / conscious-access-like representations in large language models — note: A BLIND second reading by the project on 2026-08-25: the record's own claim type/method/phenomenon were not looked at until every value here was written and substring-verified. Not applied - a person decides. Sampling frame: the 504 staged records carrying no alternatives block, so this never re-measures the 202. All 11 records with 130+ tags were taken as a census and 19 more spread one per discipline across 22 disciplines - the dense non-neuroscience cell is the one the 202 do not contain (inside them, 94-100% of the densest bands are neuroscience; among these 11, one is)., Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. consciously accessible information global workspace strength of the relation: 0.9 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span writes 'A representation becomes consciously accessible when it is posted to' the workspace; the endpoint is a variant, and operationalised-as is loose. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution. Quotation withheld because this record is not cleared for redistribution.

The full text of this article is not redistributable, so it is not included here. The evidence sentences are withheld with it; the tag layer and its grounding status remain. The article itself is at https://transformer-circuits.pub/.