Forge
markdowna5f71db5
1# Energy-First Training One-Pager
2
3## 1) Problem
4
5Current training often pays high compute cost too early:
6
7- large raw data ingestion,
8- weak stage separation,
9- expensive re-learning of already seen patterns.
10
11Goal: keep quality, reduce energy and compute.
12
13## 2) Core Hypothesis
14
15A staged protocol inspired by human development (listening -> imitation -> independent action) can reach comparable or better quality with lower total compute.
16
17## 3) Protocol (v0.1)
18
19### Phase A: Passive Incubation
20
21Model mostly observes and listens without strict output pressure.
22
23Expected effect:
24
25- builds broad priors,
26- captures environment rhythm/context,
27- lowers noise before active training.
28
29### Phase B: Imitation Micro-Actions
30
31Short supervised attempts by pattern:
32
33- repeat,
34- complete,
35- paraphrase,
36- transform.
37
38Fast and soft correction after each micro-action.
39
40Expected effect:
41
42- faster stabilization of internal representations,
43- fewer costly mistakes in later stages.
44
45### Phase C: Independent Target Tasks
46
47Focused task training with transfer from A+B.
48
49Expected effect:
50
51- less fine-tuning data needed,
52- lower expensive compute per useful capability.
53
54## 4) Experimental Design
55
56### Groups
57
58- Group A (baseline): direct training on raw data.
59- Group B: three-phase protocol (A -> B -> C).
60- Group C: A -> C (no imitation phase).
61
62### Main Metrics
63
64- quality per joule (primary),
65- steps-to-threshold quality,
66- wall-clock to threshold,
67- robustness under context shift,
68- transfer to unseen tasks.
69
70### Success Criteria
71
72- Group B quality >= Group A quality,
73- Group B uses meaningfully less compute/energy than Group A,
74- Group B outperforms Group C, confirming the value of imitation phase.
75
76## 5) Why Human Analogy Matters
77
78Humans learn motor and language interfaces through exposure + imitation + practice.
79The child hears speech before speaking, then makes uncertain attempts, then forms stable language use "on the fly".
80
81This suggests that "pre-structuring" interaction before full autonomous tasks may be compute-optimal for model learning too.
82
83## 6) Minimal v0 Execution Plan
84
851. Select one multimodal toy environment and one language-like benchmark.
862. Keep architecture fixed across groups.
873. Equalize total token/frame budget where possible.
884. Run 3 seeds per group.
895. Report not only quality, but quality-per-joule and error bars.
90
91## 7) Risks
92
93- Phase boundaries may be set poorly.
94- Benefits may depend on domain/task family.
95- Short runs can hide long-term gains or losses.
96
97Mitigation: ablations on phase duration and curriculum shape.
98
99---
100
101## RU short summary
102
103Идея: обучать по этапам, как у людей:
104
1051) сначала пассивное восприятие,
1062) потом короткие попытки по подражанию с мягкой коррекцией,
1073) и только затем целевые самостоятельные задачи.
108
109Проверка: сравнить с прямым обучением "в лоб" и вариантом без фазы подражания.
110Ключевая метрика: качество на джоуль.
111
View only · write via MCP/CIDE