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.
Model comparison — choosing between models by their evidence — the marginal likelihood and Bayes factors, the Savage–Dickey density ratio for nested models, Occam factors and the evidence framework, Bayesian model reduction and structure learning, and information criteria and free energies for singular models (ADR-026).
9 of 547 CLAIM documents. Back to the full index.
| # | Title | Summary | Status |
|---|---|---|---|
| CLAIM-128 v2 | No single ranking of translations holds across tasks: where distortion profiles trade off, only a task’s priorities rank them, and the theory of fidelity fixes no weighting | A129 §3, after the owner’s U41 “agreed.” to unifying the fidelity lineages, and restated at A151 §5, A178 Ch9.7 and A203 §22. A translation that keeps social footing and loses rhyme is not ranked above or below one that does the opposite until a task says which matters. It does not say how the components are measured, nor that they are independent. | Proposed |
| CLAIM-149 v2 | A probe family that supplies a transformation’s training signal cannot also certify what the transformation preserves: preservation must be shown under probes that were not used to train it | The owner set the frame at U51 (“a useful case study … the projected GAN work”); A184 §§3–4 argued it, with its Experiment C and closing maxim, and A203 §30 and A218 Study 6 keep it. It carries CLAIM-011’s remedy from correspondences to training objectives. Kynkäänniemi et al. (CASE-041) is a documented instance, which the exchange did not know of. The documented support is one training case, in image generation, and one selection case, in translation (CASE-071). | Active |
| CLAIM-154 v3 | A candidate structure is worth naming as a unit when its effect-to-noise ratio outruns the Occam cost of keeping it: reify when d’² exceeds the log prior-to-posterior variance ratio, where the model with the candidate stays regular at the parameters that switch it off | Dual-filed from MOM-CLM-160 (ADR-037). By the Savage–Dickey ratio, ln BF(keep) = ½(d’² − Occam), so a direction, block or sector is real enough to name exactly when δ/ε > τ with τ² = ln(s²/σ²), the information the data added about it. In this record it prices the boundary tests for a combined form: a near-decomposable block (LIT-905) earns a name when its gap beats its Occam factor. Re-checked here in the conjugate case. Narrowed to candidates whose model stays regular where they are switched off; where it is singular the rule overcharges (CLAIM-188). | Active |
| CLAIM-174 v7 | A candidate factor is an agent at a level when it is one system there, when a goal-directed description beats a device description by more than the Occam cost, and when its policy would adapt if its actions’ consequences changed; it is a willed agent when, further, a will-organization over several regimes describes it better than none, and what passes that is the jurisdiction of a will-organization | The assistant’s admission test for the owner’s thesis, filed at the owner’s request. It has three tests on a candidate factor, at a level and for a target. (1) System: it clears the threshold criterion. (2) Description: a model in which it governs its own regimes for goals beats a device model by more than the Occam cost (Orseau et al., priced as in CLAIM-154). (3) Adaptation: its governance would change if its actions’ consequences changed (Kenton et al.). It answers the circularity objection by running on candidates whose boundaries come from (1), not from agency, and it answers the permissiveness objection without the will paper’s supposition. | Proposed |
| CLAIM-188 v4 | Where the model with a candidate structure is singular at the parameters that switch the candidate off, with real log canonical threshold λ below d/2 and multiplicity m, the reification rule’s Gaussian Occam term overcharges keeping the candidate by (d/2 − λ) log n + (m − 1) log log n to leading order, so near the threshold the rule prunes candidates the Bayes factor keeps, and its derivation holds only where that model stays regular | A narrowing of CLAIM-154, filed on 2026-10-11 from a synthesis review that found the record already held the readings that limit it. CLAIM-154’s rule prices a candidate with a Gaussian Occam factor, which is exact in the conjugate case it was re-checked in. The candidates it is meant for (blocks, sectors, couplings with hidden structure) are often singular where they are switched off. There THEORY-087 puts the penalty at λ log n − (m − 1) log log n, not (d/2) log n. On CASE-152’s product model the rule prunes, near the threshold, candidates the exact Bayes factor keeps. CLAIM-154 is narrowed to models that stay regular at the switch-off. | Active |
| CLAIM-266 v2 | A family of probes that supplies an optimized system’s training signal cannot also certify what the system preserves: optimization pushes toward whatever the probes reward, including what they cannot see, so preservation must be shown under probes not used to train it | The general content of CLAIM-149, filed alone under ADR-040 so that it does not rest on CLAIM-011’s account of correspondences in translation. Training against a probe family pushes a system toward whatever the family rewards, including differences it cannot see, so the family is not a certificate of what was preserved. CASE-041 is a documented instance: a Projected FastGAN trained against ImageNet features matches StyleGAN2’s FID, also computed in ImageNet features, while doing worse in CLIP space and with human judges. | Active |
| CLAIM-307 | Retrieval pressure and evidence are different criteria for carving a category, and they move in opposite directions with copy count: evidence that a distinction exists grows with its copies, the search a bin saves shrinks with them, so the bins a deadline forces are often the ones the data least supports | From MOM-CLM-166, which reads the owner’s LEGO parable (CASE-094) against a formalism that cited it with the sign reversed. CLAIM-154 says when a distinction is real enough to name. Its evidence grows with the number of copies. The parable says when a distinction must be built to act in time: a single brick among similar ones. Expected search falls as copies grow. The two answer different questions, “is it there?” and “can it be reached in time?”, and come apart in both directions. MOM’s table was re-run here and reproduced, and a simulation of the search agrees. | Active |
| CLAIM-371 | Under the componentwise order on distortion profiles, two profiles are incomparable exactly when they trade off, and then positive weightings can rank them either way, while every positive weighting agrees with a dominance | The order-theoretic part of CLAIM-128, on its own so that it can carry its own warrant. It says nothing about translation, what the components measure, or which weighting a task fixes. Only that, under the componentwise order of TERM-043, a trade-off is incomparability, that a trade-off can be resolved either way by positive weights, and that a dominance cannot. | Active |
| CLAIM-384 | For a network linearised at its initial parameters, with an isotropic Gaussian prior of precision α on the parameter displacement and Gaussian noise of precision β, the log-determinant part of the evidence is −½ Σ log(1 + βλ/α) over the non-zero eigenvalues λ of the empirical neural tangent kernel’s Gram matrix on the training inputs, so at most n directions are priced however many parameters the network has | Active |