File size: 1,426 Bytes
34fa27c
 
ffed732
34fa27c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ffed732
34fa27c
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
cff-version: 1.2.0
message: "If you use these loss matrices, please cite the accompanying paper."
title: "Stepwise Token Learning: Token-Level Loss Trajectories Along Scaling Axes"
type: dataset
abstract: "Per-token validation-loss trajectories for one million contextualized tokens along the training-step, data-scale, and model-scale axes of LLM pre-training, plus a synthetic arithmetic example."
keywords:
  - scaling laws
  - learning-time spectrum
  - token loss
  - learning dynamics
authors:
  - family-names: Wang
    given-names: Pingjie
  - family-names: Hu
    given-names: Zechen
  - family-names: Yang
    given-names: Peiru
  - family-names: Guo
    given-names: Fu
  - family-names: Zhang
    given-names: Debing
license: MIT
url: "https://huggingface.co/datasets/applewpj/stepwise-token-learning"
repository-code: "https://github.com/applewpj/stepwise-token-learning"
preferred-citation:
  type: conference-paper
  title: "Smooth Scaling Laws Hide Stepwise Token Learning"
  url: "https://openreview.net/forum?id=THZDVZArRH"
  authors:
    - family-names: Wang
      given-names: Pingjie
    - family-names: Hu
      given-names: Zechen
    - family-names: Yang
      given-names: Peiru
    - family-names: Guo
      given-names: Fu
    - family-names: Zhang
      given-names: Debing
  collection-title: "Advances in Neural Information Processing Systems"
  conference:
    name: "NeurIPS 2026"
  year: 2026