File size: 2,874 Bytes
b4a2ae8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
{
  "schema_version": 1,
  "title": "Reproduction: High-Dimensional Learning Dynamics of Quantized Models with Straight-Through Estimator",
  "emoji": "📉",
  "space_id": "snaykey/repro-quantized-ste-dynamics",
  "paper": {
    "openreview_id": "bI9moH3UZw"
  },
  "tags": [
    "icml2026-repro",
    "paper-bI9moH3UZw"
  ],
  "updated_at": "2026-08-02T00:00:00+00:00",
  "root": {
    "slug": "index",
    "title": "Reproduction: High-Dimensional Learning Dynamics of Quantized Models with Straight-Through Estimator",
    "file": "pages/index.md",
    "children": [
      {
        "slug": "claim-1-thm43-ste-sde",
        "title": "In the high-dimensional limit, the microscopic dynamics of straight-through estimator (STE) training converge to a stochastic differential equation governed by Eq. (15) (Section IV, Theorem IV.3).",
        "file": "pages/claim-1-thm43-ste-sde/page.md",
        "children": []
      },
      {
        "slug": "claim-2-thm53-ode-concentration",
        "title": "The macroscopic summary statistics of STE training concentrate onto a deterministic ODE trajectory (Eqs. (24)-(25)) with concentration rate O_d(d^{-1/2}) as dimension d grows (Section V-B, Theorem V.3).",
        "file": "pages/claim-2-thm53-ode-concentration/page.md",
        "children": []
      },
      {
        "slug": "claim-3-plateau-drop-bitwidth",
        "title": "STE training exhibits an extended plateau followed by a sharp drop in generalization error across bit-widths b in {2,3,4,5}, with the plateau length depending on the quantization range ω (Section VI-A1/VI-A2, Figures 2-3).",
        "file": "pages/claim-3-plateau-drop-bitwidth/page.md",
        "children": []
      },
      {
        "slug": "claim-4-propvi1-input-fixedpoint",
        "title": "For input-only quantization, a closed-form stable fixed point exists under the learning-rate condition 0 < η < 2(σ_ψ^2 + λ)/σ_ψ^4 (Section VI-C1, Proposition VI.1).",
        "file": "pages/claim-4-propvi1-input-fixedpoint/page.md",
        "children": []
      },
      {
        "slug": "claim-5-thmvi3-three-regime",
        "title": "The asymptotic deviation of the STE-trained quantized model from the unquantized linear model is characterized across three regimes, with a leading-order correction term of size O_Δ(Δ^2) depending on the fractional quantization level position (Section VI-C2, Theorem VI.3).",
        "file": "pages/claim-5-thmvi3-three-regime/page.md",
        "children": []
      },
      {
        "slug": "executive-summary",
        "title": "executive-summary",
        "file": "pages/executive-summary/page.md",
        "children": []
      },
      {
        "slug": "conclusion",
        "title": "conclusion",
        "file": "pages/conclusion/page.md",
        "children": []
      }
    ]
  }
}