Topics — What a work is about. Nucleation holds whatever reading the owner files here, whether or not an Anthology of the SOTA topic could also hold it; instruction for ML practice belongs to the anthology (ADR-048). The words are seeded from the clusters the reading-time triage of 2026-09-25 found, and grow by decision.
Representation learning — how learned systems come to represent their data — self-supervised and contrastive objectives, spectral embeddings, information bottlenecks, knowledge-graph embeddings, and whether representations converge.
6 of 191 CASE documents. Back to the full index.
| # | Title | Summary | Status |
|---|---|---|---|
| CASE-005 | The language-model experiments: frame commutator, trajectory-preserving selection, ICL against TTT | A45 §5, kept in outlines v3 and v4: order effects of opposed versus independent framings in an LLM, a comparison of three ways of choosing a translation against human judgement, and prompt-only versus test- time-trained adaptation. Proposed, not run. | Active |
| CASE-020 | E3: the controlled inference harness, dry run only (C5) | Built at U25: four frame conditions (stable complicit, neutral, moralizing; shifting), 12 reset chains each, 5 generations, 240 calls, gpt-4.1-mini at temperature 0.7, seed 1729. Only a dry run was executed: “No model-generated observations were recorded.” The design is the manuscript’s Case II. | Active |
| CASE-041 v2 | FID and Projected GANs: one ImageNet-trained probe family used both to train a generator and to evaluate it | The owner pointed to FID at U50 and proposed FID with Projected GANs as a paired case study at U51. A184 framed the pair: FID evaluates a distribution through probes, and Projected GANs learn one under pressure from probes. The anthology’s readings supply the fact the exchange lacked. The Projected GAN discriminator is ImageNet-trained, FID’s features are ImageNet class evidence, and Kynkäänniemi et al. show that such a generator can match FID while doing worse in an independent feature space and with human judges. | Active |
| CASE-042 v2 | Collaborative filtering and topic models: latent relational structure recovered from interaction matrices, up to gauge, or uniquely under anchor conditions | Proposed as a laboratory by the owner at U57 and conceded as prior art after the owner’s U58. Collaborative filtering recovers a user–item matrix’s predicted relations with latent coordinates fixed only up to an invertible change; topic models recover individual topics only under nonnegativity plus separability or anchor-word conditions. It shows that recovery is possible and limited, not that a recovered component is a communicative kind. | Active |
| CASE-071 v2 | Kovacs, Deutsch and Freitag’s metric bias in MBR decoding: translations a metric selects, that the metric and its neighbours rate better and human raters do not | Minimum Bayes risk and quality-estimation decoding pick, from 128 sampled translations, the one a neural metric scores best. That metric and other neural metrics rate the picks significantly better than greedy decoding; human MQM raters find them no better, and for four system and language-pair combinations significantly worse. CASE-041’s result in text, by selection rather than training. The distinction lost is accuracy, not a pragmatic one; a stipulated English-to-Russian ты/вы variant states the design that would make it pragmatic. | Active |
| CASE-187 | A proposed study: agents matched on competence over a declared family of goals, one planning with an explicit transition model, one a directly learned goal-conditioned policy, one a distributed predictive-state controller, compared on the original goals, on on-demand predictive and interventional queries, under targeted internal interventions, and after environmental change and damage to their predictive machinery, to find which world-modeling achievements are linked and which come apart under matched resources | Proposed by the owner at the second review of the filing of the exchange “Analyze nucleation concepts” (2026-10-31), as the experiment that would sharpen the world-model distinction. Not run; its architectures, tasks and margins are left to a preregistration. | Active |