Paper

The Weight Norm Sets the Grokking Timescale: A Causal Delay Law

Unreadunread

AI summary

arXiv:2606. 13753v1 Announce Type: new Abstract: Grokking is the delayed onset of generalization in neural networks, arising long after they fit the training data.

Reading status

Priority

About this paper

Authors: Truong Xuan Khanh, Doan Hoang Viet, Luu Duc Trung, Phan Thanh Duc
Released: 2026-06-15 (2mo ago)
Topic: Related to current project
Relation to my work: Read next because The Weight Norm Sets the Grokking Timescale: A Causal Delay Law overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: under, alpha, rate, without. Source: arxiv cs.LG (Machine Learning).
arXiv: 2606.13753

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to The Weight Norm Sets the Grokking Timescale: A Causal Delay Law. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.