meta-architect-discoveries

maxie-12321/meta-architect-discoveries

meta-architect-discoveries is a machine learning model from maxie-12321. Use it for the machine learning task on the model card, and read the license before you ship it in a product.

Last updated: 2026-02-17 00:05:03 UTC Upload 9 | Tracked: 50 | Best reward: 11.28

Downloads · 30 days

Access

Public

Updated Feb 17, 2026

Repo size

1 GB

Likes

Public

Hugging Face

Repo makeup

Click a slice to open those files.

.pt1 GB · 100%

At a glance

Access: Public
Created: Feb 16, 2026
Updated: Feb 17, 2026
SHA: 783e1818

Model card

From the Hugging Face model README

Meta-Architect Discoveries

Last updated: 2026-02-17 00:05:03 UTC Upload #9 | Tracked: 50 | Best reward: 11.28

Top Architectures

Rank	Reward	Acc	Gen	Eff	Spd	CL	Params	Description
1	11.28	0.928	0.523	0.832	0.953	0.883	55,306	`reg none → norm group → dim 64 → struct split merge → act id`
2	11.10	0.953	0.566	0.545	0.968	1.000	135,562	`head double → act gelu → norm batch → struct repeat 3 → act`
3	10.92	0.891	0.541	0.713	0.953	0.966	67,946	`norm group → norm batch → dim 32 → norm group → act sigmoid`
4	10.87	0.953	0.564	0.525	0.933	0.966	146,122	`norm rms → layer gated → act square → layer bottleneck → dim`
5	10.70	0.931	0.538	0.513	0.951	1.000	146,730	`norm group → layer dual path → act silu → norm none → layer`
6	10.69	0.940	0.533	0.514	0.959	1.000	148,746	`e blk 182 → norm batch → e blk 163 → norm layer → act elu →`
7	10.63	0.944	0.528	0.519	0.929	0.989	147,114	`norm batch → e blk 174 → head simple → dim 48 → struct widen`
8	10.59	0.918	0.521	0.527	0.922	1.000	135,050	`layer residual → e act 168 → norm group → layer bilinear → a`
9	10.52	0.946	0.569	0.489	0.959	1.000	166,426	`comb mul → norm batch → e blk 187 → struct skip connect → st`
10	10.36	0.915	0.523	0.520	0.921	0.892	137,677	`e blk 176 → comb norm mul → dim 32 → act softsign → act sign`
11	10.29	0.947	0.567	0.449	0.946	1.000	197,837	`dim 96 → act silu → act identity → act identity → layer trip`
12	10.25	0.948	0.550	0.463	0.940	1.000	185,770	`layer gated → act silu → comb norm mul → norm group → reg dr`
13	10.22	0.945	0.565	0.443	0.923	1.000	202,365	`m act 180 → layer gated → layer triple mix → norm group → la`
14	10.22	0.954	0.573	0.629	0.940	0.552	102,026	`act abs → norm batch → act gelu → comb add → reg none`
15	10.15	0.936	0.536	0.481	0.958	0.963	168,332	`norm group → layer hadamard → norm none → layer mixture → ac`
16	10.11	0.946	0.542	0.445	0.957	1.000	201,099	`comb chain → norm batch → comb chain → reg dropout med → e b`
17	10.07	0.947	0.574	0.427	0.945	0.914	218,253	`reg dropout med → act softsign → act gelu → act silu → e blk`
18	10.07	0.928	0.537	0.490	0.919	0.924	159,549	`dim 32 → struct repeat 2 → norm layer → dim 48 → layer tripl`
19	10.02	0.927	0.525	0.455	0.948	1.000	184,586	`struct reverse dim → act tanh → comb mul → struct repeat 3 →`
20	9.99	0.936	0.582	0.369	0.949	1.000	285,483	`layer gated → layer bottleneck → comb lerp → dim 512 → dim 1`

Token Sequences (for reproduction)

#1 (R=11.28): [83, 74, 69, 25, 92, 68, 113, 21, 110, 143, 33, 155, 136, 136, 10, 12, 7, 118, 92, 130, 124]

#2 (R=11.10): [121, 14, 70, 111, 16, 120]

#3 (R=10.92): [77, 76, 74, 70, 75, 102, 106, 102, 90, 46, 74, 123, 12, 82, 12, 48, 30, 146, 30, 10, 51, 79, 77, 180, 66, 83, 104]

#4 (R=10.87): [126, 73, 141, 51, 101, 17, 142, 55, 47, 78, 92, 26, 13, 109, 108, 112, 151, 38, 127, 83, 101, 30, 13, 84, 128, 85, 157, 33, 112, 30, 14, 104, 81, 84, 74, 149, 24, 113, 131, 123, 124, 10, 91, 68]

#5 (R=10.70): [108, 154, 74, 146, 140, 62, 128, 13, 72, 26, 140, 41, 51, 133, 74, 91, 136, 126]

#6 (R=10.69): [136, 182, 70, 129, 163, 71, 29, 20, 7, 11, 103, 124, 46, 63, 78, 70, 111, 70, 124, 68]

#7 (R=10.63): [136, 70, 79, 174, 41, 120, 91, 116, 104]

#8 (R=10.59): [75, 53, 168, 75, 74, 39, 52, 135, 21, 37, 157]

#9 (R=10.52): [42, 31, 70, 129, 137, 128, 187, 112, 27, 154, 112, 3, 91, 24, 102, 39, 53, 103, 14, 82, 113, 163, 176, 47, 70, 70, 26, 41, 70, 127, 31]

#10 (R=10.36): [176, 37, 90, 24, 18, 118, 70, 126, 133, 78, 146, 23, 137, 84, 147, 78, 90, 88, 53, 73, 146, 145, 82, 91, 138, 92, 87, 105, 41, 8, 138, 154, 128, 112, 176, 118, 5, 79, 83, 14, 147, 29, 38, 112, 171, 136, 147, 69, 57, 154, 8, 47, 91, 63, 8, 168, 90, 150, 41, 137, 14, 89, 131, 102, 110, 151, 47, 112, 19, 176, 99, 16, 132]

Created: Feb 16, 2026
Updated: Feb 17, 2026