finding

active

finding:under-spatio-permutation-controls-two-cases-layer-32-of-mixtral-8x7b-on-strange-stories-iit-4-0-linguistic-spans-entire-and-complement-satisfy-all-three-criteria

Under spatio permutation controls, two cases (Layer 32 of Mixtral-8x7B on Strange Stories, IIT 4.0, Linguistic Spans: Entire and Complement) satisfy all three criteria.

Contrasts with temporal permutation results; constitutes the most suggestive evidence of potential consciousness phenomena in LLM representations.

Source paper

extracted_from

Can "consciousness" be observed from large language model (LLM) internal states? Dissecting LLM representations obtained from Theory of Mind test with Integrated Information Theory and Span Representation analysis

(2025) · Li, Jingkai

Neighborhood — ranked by edge-count

Claims (1)

claim

LLM representations exhibit intriguing patterns under spatio-permutational analyses, suggesting a potentially profound yet tentative indication of consciousness.
supports
Qualified positive claim from spatio permutation analysis where two cases satisfy all three criteria.

Hypotheses (1)

hypothesis

We hypothesize that 'consciousness' phenomena can be observed in the internal states of an LLM, specifically in its learned representations when analyzed as a sequence.
associated_withsupports
Primary research hypothesis driving the entire study; operationalized via three criteria.

Related by similarity (8)

cosine ≥ 0.65 · no typed edge

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

Under spatio permutation controls, IIT consciousness estimates outperform Span Representation in mean AUC in several cases (LLaMA3.1-70B on Hinting and Irony, Mistral-7B on Irony, LLaMA3.1-8B on Strange Stories).finding0.798
Contrasts with temporal permutation where Span Representation dominates; suggests spatio permutation reveals different dynamics.
None of the cases identified under temporal permutation satisfy the Criterion 1 threshold of >80% 'good' cases for any ToM task.finding0.795
Even the rare cases where good > bad do not reach the 80% significance threshold required by Criterion 1.
Layer 29 (indexed at 10) of LLaMA3.1-8B on Strange Stories (2 scores) satisfies Criteria 1 and 2 under IIT 4.0 (temporal permutation).finding0.787
Third promising case from temporal permutation analysis.
Layer 24 (indexed at 8) of LLaMA3.1-8B on Hinting satisfies Criteria 1 and 2 under both IIT 3.0 and IIT 4.0 (temporal permutation).finding0.778
One of the most promising cases; approximately corresponds to the 2/3 layer of LLaMA3.1-8B.
Under temporal permutation control, no cases meeting all three criteria for observed 'consciousness' phenomenon were found among the 165,365 valid samples.finding0.775
Primary negative result of the study: temporal permutation analysis finds no statistically significant indicators of consciousness in LLM representations.
All cases satisfying Criteria 1 and 2 (two out of three) originate from deeper transformer layers and/or the 2/3 layer of LLMs.finding0.771
Consistent with literature that deeper layers encode semantic information and align with human brain activity.
Several Mixtral-8x7B samples could not be initialized as valid networks using PyPhi under IIT 4.0 and were excluded.finding0.762
Methodological limitation disproportionately affecting the largest MoE model, constraining generalizability.
Adding a single disambiguating example (12−9=21) aligns divergent M1-M4 interpretations under tested seedsfinding0.754
E1 finding consistent with threshold-crossing: near-threshold state resolved by one additional anchor