11. What I could NOT verify (be explicit with stakeholders)

Chalmers, "Could a Large Language Model be Conscious?" (2303.07103) — body text. No arXiv HTML, no ar5iv rendering (both errored), PDF rejected by the fetcher, Boston Review page returned only its title, and Chalmers' own site does not host the paper (consc.net/papers/llm.pdf → 404; absent from his AI-and-computation and consciousness index pages). I have only the abstract. His named obstacles (recurrent processing, a global workspace, unified agency), his two-part conclusion, and the quoted wording are verified. His fuller list of requirements, and any detailed argument he gives, are NOT verified. Do not attribute specific further requirements to him without fetching the PDF directly.
Chalmers' "easy problems" list. Facing Up to the Problem of Consciousness is PDF-only (consc.net/papers/facing.pdf) and I could not read it. I did not restate the list from memory.
The unfolding argument's full text. Abstract verified via OpenAlex; no arXiv preprint exists (and 1810.11143 is a different paper). The precise formal statement of the unfolding construction is taken from the abstract and title only.
The 2023 pseudoscience preprint's body. Title, DOI, year, first-author pseudonym, named co-authors and abstract verified. The OSF page returned only "OSF". Full signatory list, argument structure, and submission status unverified.
Tsuchiya, Andrillon & Haun's reply to the unfolding argument. Title and DOI only; no abstract available via OpenAlex. Their counter-argument is unverified.
Lau & Rosenthal 2011, and Wilterson & Graziano PNAS 2021. Citations, titles, venues, years, authorship verified via OpenAlex; abstracts not available, so no quotes.
Aguilera, Millidge, Tschantz & Buckley, "How particular is the physics of the free energy principle?" Referenced by its title in four published replies I did retrieve, but I did not verify its own arXiv ID, so I have deliberately not stated one.
Kleiner & Hoel, "Falsification and consciousness." Referenced by name in the title of a reply I did retrieve; its own ID is unverified, so I have not stated one.
A paper titled "The Free Energy Principle: A Critique" by Colombo & Wright. I could not verify that this exists. What I verified is Colombo & Wright's Synthese paper on first principles in the life sciences (doi 10.1007/s11229-018-01932-w). The title in the brief may be a conflation.
Fleming & Lau, "How to measure metacognition" (doi 10.3389/fnhum.2014.00443). Citation verified; full text not retrieved (the Frontiers page returned only its shell). My account of meta-d' is derived from the applied 2025–2026 papers that cite and use it, not from Fleming & Lau directly.
Butlin et al.'s full paper beyond §1–2. I retrieved the front matter, executive summary, the method section, and Table 1 via ar5iv; the fetch was truncated partway through §1.2.3. Their §3 (how each indicator could be implemented) and §3.2 (case studies of specific systems) are not verified in detail.
IIT 4.0's handling of falsifiability. The 53-page body was not retrievable (PDF only; ar5iv failed). Whether IIT 4.0 explicitly answers the unfolding argument is unverified.
Doerig et al., "Hard criteria for empirical theories of consciousness" — the 13 criteria themselves. Abstract verified; the checklist is in the PDF, which I could not read.
The meta-d' formula and its assumptions. I state it as "type-2 SDT measure in type-1 d' units" based on how the 2025–2026 applied papers use it. The formal derivation and its known limitations (e.g. the requirement for a 2AFC type-1 decision, and the meta-I₂ᵣ fallback for open-ended tasks) are from Cacioli arXiv 2603.25112, not from the primary methodology paper.
Anything behind a PDF-only wall. This includes, in addition to the above: Chalmers 1995; Tononi's original Φ derivations beyond the 2004 BMC abstract; the full IIT 4.0 postulates; Lau & Rosenthal 2011; Wilterson & Graziano 2021; the Synthese Colombo & Wright body; and the BBS Bruineberg et al. body. For all of these I have abstracts and/or titles only, and I have marked them accordingly throughout.

#One caution about the corpus

A large number of the arXiv items I retrieved carry 2026 timestamps and appear at AAAI 2026 Spring Symposium ("Machine Consciousness: Integrating Theory, Technology, and Philosophy"), EMNLP '26, CogSci 2026, and in Neuroscience of Consciousness 2026. These are live in-world publications returned by the arXiv API, and I have cited them as such. Several are single-author preprints with no independent replication (e.g. Phua 2512.19155, Xie 2604.11914, Sanyal 2602.23232, Servajean 2603.29693). Their protocols are the valuable part and are worth copying; their findings should be treated as preliminary until replicated. The one large-N, multi-lab, pre-registered result in this report is the Nature 2025 Cogitate adversarial collaboration.