{"slug": "response-relevant-information-is-not-sufficient-for-neural-training-state", "title": "Response-Relevant Information Is Not Sufficient for Neural Training-State Equivalence", "summary": "A study posted to arXiv (2611.23245v1) finds that response-relevant information alone is not sufficient evidence that two neural networks occupy the same training state. On fresh runs of a one-layer transformer learning modular addition, the authors measured highly reproducible responses to a weight-decay pulse relative to between-checkpoint differences (rho_state = 0.934 at 500 steps), but the apparent 16.1% advantage of fifteen weight-spectrum, representation-geometry and optimiser-geometry descriptors at matching checkpoints by future response did not separate from a conditional placebo preserving each descriptor's fitted dependence on conventional observables. The authors conclude that information a decoder can extract need not be organised in the pairwise geometry that similarity judgments require.", "body_md": "arXiv:2611.23245v1 Announce Type: new \nAbstract: Response-relevant information is not sufficient evidence that two neural networks are in the same training state. We define training-state equivalence operationally: two checkpoints are equivalent, relative to an intervention family and horizon, if they respond alike when training is perturbed and continues. On fresh runs of a one-layer transformer learning modular addition, response to a weight-decay pulse is highly reproducible relative to between-checkpoint differences (rho_state = 0.934 at 500 steps), establishing the target as measurable. Fifteen descriptors of weight spectra, representation geometry and optimiser geometry carry within-run response-relevant information beyond weight norm and loss on a development dataset, predicting held-out response deviation across seed blocks. But their apparent 16.1% advantage at matching checkpoints by future response does not separate from a conditional placebo preserving each descriptor's fitted dependence on the conventional observables, and a calibrated positive control makes simple insensitivity an unlikely explanation. Information a decoder can extract need not be organised in the pairwise geometry that similarity judgments require.", "url": "https://wpnews.pro/news/response-relevant-information-is-not-sufficient-for-neural-training-state", "canonical_source": "https://arxiv.org/abs/2611.23245", "published_at": "2026-11-19 05:00:00+00:00", "updated_at": "2026-11-19 05:12:33.861706+00:00", "lang": "en", "topics": ["ai-research", "machine-learning", "neural-networks"], "entities": ["arXiv", "one-layer transformer"], "also_reported_by": [], "takeaway": "Checkpoint similarity judged by response-relevant descriptors can be an artifact: fifteen descriptors' 16.1% edge in matching checkpoints by future response vanished against a conditional placebo, so validate such metrics against a placebo before trusting them.", "recommendation": {"advice": "When using representational or weight-spectrum descriptors to match checkpoints, test them against a conditional placebo that preserves each descriptor's dependence on loss and weight norm before relying on the apparent advantage.", "playbook": "evals-testing", "playbook_url": "https://wpnews.pro/playbooks/evals-testing"}, "alternates": {"html": "https://wpnews.pro/news/response-relevant-information-is-not-sufficient-for-neural-training-state", "markdown": "https://wpnews.pro/news/response-relevant-information-is-not-sufficient-for-neural-training-state.md", "text": "https://wpnews.pro/news/response-relevant-information-is-not-sufficient-for-neural-training-state.txt", "jsonld": "https://wpnews.pro/news/response-relevant-information-is-not-sufficient-for-neural-training-state.jsonld"}}