CORTEXA
← Browse
arxivcs.LGcs.AI2026-07-06

Grokking Is Conditional and Fragile: A Fully-Tractable, Multi-Seed Study at 12K Parameters

Yoshiyuki Ootani

Grokking -- the delayed onset of generalization long after a network has fit its training set - -is usually studied in models too large to read completely and reported from single training runs. We instead study a publicly released ~11,856-parameter Llama-style transformer (Glimmer-1-Base) on modular arithmetic, small enough to enumerate its weights, attention, and full input-output map, and we measure grokking as a multi-seed rate rather than a single outcome. In this fully-tractable regime grokking is a conditional, fragile phase transition. It is gated by training-set coverage, whose threshold tracks output cardinality (the modulus) more than task structure, an ordering that holds above the transition and across a ten-fold change in domain size. Weight decay reproduces the Omnigrok inverted-U at 12K parameters, a positive control on the rate measurement. Grokking also sits on a numerical knife-edge: two perturbations of the floating-point environment -- CPU thread count (reduction order) and CPU-versus-GPU execution -- each flip a minority of same-seed outcomes without a detectable shift in the aggregate rate. Decomposition into sub-task specialists helps chiefly by making coverage cheap rather than by adding supervision. Methodologically, multi-seed control under a fixed numerical environment overturns three dramatic single-run narratives in our own data, each a seed confound. The unit of evidence for grokking must therefore be a multi-seed rate under a pinned numerical environment, checked where possible against a direct reading of the model.

View free PDFSource page

Related papers

arxivcs.LGcs.AIeess.SY2026-06-30

Preference-Conditioned Multi-Objective Reinforcement Learning for Runtime-Tunable Transit Signal Priority

Philip-Roman Adam, Stefanie Schmidtner

Transit signal priority (TSP) requires balancing competing objectives: reducing bus delay while limiting adverse impacts on non-bus traffic and avoiding extreme waits for a subset of vehicles. Existing reinforcement-learning (RL) approaches to TSP typically encode transit-aware f…

View free PDFSource page
arxivcs.LGcs.AI2026-07-06

The Active Ingredient in Muon's Grokking

Yufeng Wang

The Muon optimizer reaches the grokking threshold on modular arithmetic faster than AdamW. Prior work attributes this to "spectral-norm constraints plus orthogonalized momentum" but does not isolate which mechanism matters. To better understand Moun's behavior, we run multi-seed…

View free PDFSource page
arxivcs.AIcs.LG2026-06-30

Self-Study Reconsidered: The Hidden Fragility of Learning from Self-Generated QA

Ekaterina Alimaskina, Denis Shveykin, Gleb Molodtsov, Igor Shalygin, Alexey Kadeishvili, Aleksandr Beznosikov

Language models are increasingly taught from synthetic question--answer (QA) supervision: a model generates questions about a document, answers them from the same text, and the resulting pairs are used to fine-tune, distill, or compress knowledge into another model. We show that…

View free PDFSource page
arxivcs.AIcs.LGcs.MA2026-06-29

Why Solve It Twice? Hierarchical Accumulation of Skills for Transfer-Efficient ML Engineering

Yongbin Kim, Yashar Talebirad, Osmar R. Zaiane

ML engineering agents waste compute rediscovering known techniques because every competition is a cold start. We present HASTE, a hierarchical multi-agent system that organizes cross-competition knowledge into three scope tiers (global, domain, and competition-specific), each cou…

View free PDFSource page
arxivcs.LGcs.AIcs.IT2026-07-17

Capacity and Redundancy Trade-offs in Multi-Task Learning

Asif Khan

In multi-task learning (MTL) negative transfer is often considered as an optimization artifact, but it can also be viewed as a consequence of limited shared capacity and weak task redundancy. We investigate this effect through a Capacity--Redundancy (CR) identity that decomposes…

View free PDFSource page