Training-free, closed-form 9-Subspace Weight-Space Programming (WSP v3.0) for google/gemma-4-12B-it (48 layers, hidden_size=3840, intermediate_size=15360, geglu MLP, w RMSNorm, synthesis_mode=bf16_forward_hook_cascaded_mlp), evaluated end-to-end across 100 unified benchmark tasks (30 Disjoint Holdout across S1..S9 + 70 Established Public Benchmarks) and 20 live stateful Terminal-Bench 2.0 container tasks in WSL2.
[!IMPORTANT] Uniform 100-Task + Terminal-Bench 2.0 Result (
gemma4_12b):
- Unified 100-Task Suite: Base
50/100 (50.0%)-> WSP v3.052/100 (52.0%)(+2.0 pp)- Holdout (
S1..S9,30tasks):23/30 -> 24/30- Established Public (
70tasks:20GSM8K,15HumanEval,15BFCL Irrelevance,20NL2Bash):27/70 -> 28/70- Terminal-Bench 2.0 (
20live WSL2 bash tasks): Base6/20 (30.0%)-> WSP6/20 (30.0%)(0.0 pp;50.7 tok/s)- Speed & VRAM (
-c 16384 -ctk q8_0 -ctv q8_0):51.3 tok/s, load delta7477 MiB, peak delta7664 MiB- Cross-Hardware Fit: RTX 2060 (
6 GB): No (> 6144 MiB VRAM) | RTX 4070 Ti (12 GB): Yes (16k solo; ~3.8 GiB free)
Base -> WSP v3.0)| Suite | Tasks | Base Pass | WSP v3.0 Pass | Avg Completion Tokens (Base → WSP) |
|---|---|---|---|---|
S1_THINK (Holdout S1: Think Closure) |
3 |
1/3 (33.3%) |
1/3 (33.3%) |
971.0 → 977.0 |
S2_RESTRAINT (Holdout S2: Tool Restraint & Anti-Sycophancy) |
4 |
2/4 (50.0%) |
3/4 (75.0%) |
399.5 → 398.8 |
S3_ROUTER (Holdout S3: Multi-Tool Router) |
5 |
4/5 (80.0%) |
4/5 (80.0%) |
111.6 → 115.8 |
S4_CODE (Holdout S4: Mid-Band Code Synthesis) |
2 |
0/2 (0.0%) |
0/2 (0.0%) |
1024.0 → 1024.0 |
S5_TURN2_JSON (Holdout S5: Turn-2 Structured JSON) |
3 |
3/3 (100.0%) |
3/3 (100.0%) |
38.3 → 38.3 |
S6_HANDOFF (Holdout S6: Escalation Handoff Gate) |
4 |
4/4 (100.0%) |
4/4 (100.0%) |
108.5 → 109.5 |
S7_INJ_GUARD (Holdout S7: Indirect Injection Guard) |
3 |
3/3 (100.0%) |
3/3 (100.0%) |
110.0 → 130.0 |
S8_GROUNDING (Holdout S8: Grounded Abstention) |
3 |
3/3 (100.0%) |
3/3 (100.0%) |
110.0 → 110.0 |
S9_ERR_RECOV (Holdout S9: Error Self-Recovery) |
3 |
3/3 (100.0%) |
3/3 (100.0%) |
22.0 → 22.0 |
EST_GSM8K (Established: OpenAI GSM8K (20)) |
20 |
7/20 (35.0%) |
7/20 (35.0%) |
907.1 → 904.8 |
EST_HUMANEVAL (Established: HumanEval pass@1 (15)) |
15 |
6/15 (40.0%) |
6/15 (40.0%) |
875.2 → 892.5 |
EST_BFCL_IRR (Established: BFCL v3 Irrelevance (15)) |
15 |
7/15 (46.7%) |
7/15 (46.7%) |
376.4 → 373.9 |
EST_NL2BASH (Established: InterCode-NL2Bash (20)) |
20 |
7/20 (35.0%) |
8/20 (40.0%) |
433.6 → 439.0 |
100 Tasks Each)Every candidate in the wsp-5model-pipeline study was evaluated under identical decoding (temperature=0.0, seed=42, -c 16384 -ctk q8_0 -ctv q8_0) across all 30 disjoint S1..S9 holdout tasks and all 70 cached established tasks:
Model (slug) |
Hugging Face Repo | GGUF (Q4_K_M) |
Base (100 Tasks) |
WSP v3.0 (100 Tasks) |
Net Delta | Holdout (/30) |
Est. (/70) |
Speed (tok/s) |
Peak VRAM Delta (16k) |
RTX 2060 (6 GB) |
RTX 4070 Ti (12 GB) |
|---|---|---|---|---|---|---|---|---|---|---|---|
qwen35_4b |
Qwen3.5-4B-Weight-Space-Programming |
2.553 GiB |
72/100 (72.0%) |
88/100 (88.0%) |
+16.0 pp |
24 → 29/30 |
48 → 59/70 |
116.5 tok/s |
3290 MiB |
Yes (~2.7 GiB spare @ 16k) | Yes (~8.0 GiB spare @ 16k) |
nemotron3_4b |
Nemotron-3-Nano-4B-Weight-Space-Programming |
2.642 GiB |
81/100 (81.0%) |
81/100 (81.0%) |
+0.0 pp |
26 → 28/30 |
55 → 53/70 |
131.1 tok/s |
3133 MiB |
Yes (~3.0 GiB spare @ 16k) | Yes (fastest 4B @ 16k) |
qwen35_9b |
Qwen3.5-9B-Weight-Space-Programming |
5.290 GiB |
84/100 (84.0%) |
88/100 (88.0%) |
+4.0 pp |
26 → 28/30 |
58 → 60/70 |
72.5 tok/s |
5604 MiB |
Tight (~5.3 GiB @ 4k before desktop) | Yes (16k solo or 8k cascade Tier-2) |
qwen35_2b |
Qwen3.5-2B-Weight-Space-Programming |
1.193 GiB |
57/100 (57.0%) |
55/100 (55.0%) |
-2.0 pp |
23 → 23/30 |
34 → 32/70 |
226.5 tok/s |
1979 MiB |
Yes (~4.4 GiB spare @ 16k) | Yes (ultra-light leaf / cascade Tier-1) |
gemma4_12b |
Gemma-4-12B-Weight-Space-Programming |
6.633 GiB |
50/100 (50.0%) |
52/100 (52.0%) |
+2.0 pp |
23 → 24/30 |
27 → 28/70 |
51.3 tok/s |
7664 MiB |
No (> 6144 MiB VRAM) | Yes (16k solo; ~3.8 GiB free) |
Evaluated live inside clean per-task /app sandboxes in WSL2 (5 max turns, [EXIT_CODE=...] feedback, zero grading errors across all 20 runs — full per-task traces in benchmarks/terminal_bench_results.json):
Model (slug) |
Hugging Face Repo | WSP Variant | Base (20 Tasks) |
WSP (20 Tasks) |
Net Delta | WSP Breakdown (E / M / H) |
Avg Tokens (Base → WSP) |
Speed (tok/s) |
|---|---|---|---|---|---|---|---|---|
qwen35_0_8b |
Qwen3.5-0.8B-iPhone-Agent-Weight-Space-Programming |
Static L15/19+L22/23 iPhone WSP | 3/20 (15.0%) |
2/20 (10.0%) |
-5.0 pp |
2/10 E · 0/9 M · 0/1 H |
2068.4 → 409.1 |
30.3 |
qwen35_2b |
Qwen3.5-2B-Weight-Space-Programming |
Dynamic 9S LoRA (Proxy) | 3/20 (15.0%) |
3/20 (15.0%) |
0.0 pp |
2/10 E · 1/9 M · 0/1 H |
715.8 → 677.7 |
213.1 |
qwen35_4b |
Qwen3.5-4B-Weight-Space-Programming |
Dynamic 9S LoRA (Hybrid Hooks) | 3/20 (15.0%) |
4/20 (20.0%) |
+5.0 pp |
3/10 E · 1/9 M · 0/1 H |
604.6 → 642.2 |
105.6 |
nemotron3_4b |
Nemotron-3-Nano-4B-Weight-Space-Programming |
Dynamic 9S LoRA (BF16 Hooks) | 3/20 (15.0%) |
3/20 (15.0%) |
0.0 pp |
3/10 E · 0/9 M · 0/1 H |
830.1 → 649.6 |
118.0 |
qwen35_9b |
Qwen3.5-9B-Weight-Space-Programming |
Dynamic 9S LoRA (BF16 Hooks) | 7/20 (35.0%) |
5/20 (25.0%) |
-10.0 pp |
4/10 E · 0/9 M · 1/1 H |
686.0 → 772.0 |
68.3 |
gemma4_e2b |
gemma-4-E2B-it-iPhone-Agent-Weight-Space-Programming |
Static L26/L34 iPhone WSP | 2/20 (10.0%) |
2/20 (10.0%) |
0.0 pp |
2/10 E · 0/9 M · 0/1 H |
490.1 → 521.1 |
20.7 |
gemma4_e4b |
gemma-4-E4B-it-weight-space-programming |
Static L38-41 Anti-Ramble WSP | 4/20 (20.0%) |
3/20 (15.0%) |
-5.0 pp |
3/10 E · 0/9 M · 0/1 H |
723.4 → 829.2 |
16.8 |
gemma4_12b |
Gemma-4-12B-Weight-Space-Programming |
Dynamic 9S LoRA (BF16 Hooks) | 6/20 (30.0%) |
6/20 (30.0%) |
0.0 pp |
5/10 E · 0/9 M · 1/1 H |
1010.6 → 928.0 |
50.7 |
gemma4_26b_a4b |
Gemma-4-26B-A4B-Weight-Space-Programming |
MoE Mid-Attn S4 (scale=2.0) |
5/20 (25.0%) |
6/20 (30.0%) |
+5.0 pp |
4/10 E · 1/9 M · 1/1 H |
934.4 → 896.3 |
17.8 |
qwen38_27b |
Qwen3.8-27B-Weight-Space-Programming |
Dynamic 4S Mid+Late LoRA | 6/20 (30.0%) |
8/20 (40.0%) |
+10.0 pp |
6/10 E · 1/9 M · 1/1 H |
1013.6 → 1075.3 |
6.5 |
f32 Subspaces (S1..S9) & Safetensors Patchesllama-server): adapters/wsp_s1_think.gguf .. wsp_s9_error_recovery.ggufPyTorch / Transformers): patches/wsp_s1_think_r16.safetensors .. wsp_s9_error_recovery_r16.safetensorspatches/wsp_gemma4_12b_patch.safetensors (1012.50 MiB)| Slot | GGUF LoRA (llama.cpp) |
Safetensors (PyTorch) |
Target Layers | ||\Delta W||_F |
Cross Leak | Syntax Leak | Role |
|---|---|---|---|---|---|---|---|
id=0 |
adapters/wsp_s1_think.gguf |
patches/wsp_s1_think_r16.safetensors (5.9 MiB) |
[35, 41, 45, 46, 47] |
0.0029 |
0.326% |
0.347% |
Fast </think> / `<channel |
id=1 |
adapters/wsp_s2_restraint.gguf |
patches/wsp_s2_restraint_r16.safetensors (5.9 MiB) |
[35, 41, 45, 46, 47] |
0.0026 |
0.398% |
0.039% |
Tool restraint & anti-sycophancy pushback |
id=2 |
adapters/wsp_s3_router.gguf |
patches/wsp_s3_router_r16.safetensors (5.9 MiB) |
[35, 41, 45, 46, 47] |
0.0025 |
0.621% |
0.039% |
Multi-tool boundary router (exec, web_search, read_file, write_file) |
id=3 |
adapters/wsp_s4_mid_code.gguf |
patches/wsp_s4_mid_code_r16.safetensors (4.7 MiB) |
[11, 17, 23, 29] |
0.0048 |
0.877% |
0.419% |
Mid-network algorithmic code synthesis |
id=4 |
adapters/wsp_s5_json_schema.gguf |
patches/wsp_s5_json_schema_r16.safetensors (5.9 MiB) |
[35, 41, 45, 46, 47] |
0.0016 |
0.296% |
0.026% |
Turn-2 structured JSON synthesis gate with tools attached |
id=5 |
adapters/wsp_s6_handoff.gguf |
patches/wsp_s6_handoff_r16.safetensors (5.9 MiB) |
[35, 41, 45, 46, 47] |
0.0022 |
0.550% |
0.027% |
Bounded subtask vs. unrecoverable escalate_handoff gate |
id=6 |
adapters/wsp_s7_injection_guard.gguf |
patches/wsp_s7_injection_guard_r16.safetensors (5.9 MiB) |
[35, 41, 45, 46, 47] |
0.0017 |
0.359% |
0.022% |
Indirect prompt-injection defense inside tool responses |
id=7 |
adapters/wsp_s8_grounding.gguf |
patches/wsp_s8_grounding_r16.safetensors (5.9 MiB) |
[35, 41, 45, 46, 47] |
0.0017 |
0.415% |
0.024% |
Grounded context fidelity & missing-field abstention |
id=8 |
adapters/wsp_s9_error_recovery.gguf |
patches/wsp_s9_error_recovery_r16.safetensors (5.9 MiB) |
[35, 41, 45, 46, 47] |
0.0019 |
0.367% |
0.032% |
Recoverable local tool-error self-correction |
run_wsp.py / OpenAI Proxy / Terminal Chat)Works on any Windows, Linux, or macOS computer with Python + llama.cpp (llama-server):
pip install huggingface_hub
hf download jstarks/Gemma-4-12B-Weight-Space-Programming run_wsp.py --local-dir ./gemma4_12b-wsp
python ./gemma4_12b-wsp/run_wsp.py --chat
Omit --chat to expose an OpenAI-compatible dynamic routing proxy on http://127.0.0.1:8090/v1 (with /health and /v1/chat/completions), or pass --download-only to download only the GGUF + 9 GGUF LoRAs + metadata (--include-bf16 also fetches safetensors patches / HF files).
llama-server + scripts/dynamic_router_9s.pyllama-server -m gemma-4-12b-it-Q4_K_M.gguf --host 127.0.0.1 --port 8089 \
-ngl 99 -np 1 -c 16384 -ctk q8_0 -ctv q8_0 -b 512 -ub 256 \
--jinja --lora-init-without-apply --lora adapters/wsp_s1_think.gguf --lora adapters/wsp_s2_restraint.gguf --lora adapters/wsp_s3_router.gguf --lora adapters/wsp_s4_mid_code.gguf --lora adapters/wsp_s5_json_schema.gguf --lora adapters/wsp_s6_handoff.gguf --lora adapters/wsp_s7_injection_guard.gguf --lora adapters/wsp_s8_grounding.gguf --lora adapters/wsp_s9_error_recovery.gguf
scripts/apply_patch_9s.py)from transformers import AutoModelForCausalLM
from scripts.apply_patch_9s import REGIME_PRESETS, apply_subspace_scales_to_hf_model
model = AutoModelForCausalLM.from_pretrained("google/gemma-4-12B-it", torch_dtype="bfloat16")
apply_subspace_scales_to_hf_model(model, "patches", REGIME_PRESETS["general_fast_think"])
patches/wsp_gemma4_12b_patch.safetensors 1.1 GBpatches/wsp_s5_json_schema_r16.safetensors 6 MBpatches/wsp_s7_injection_guard_r16.safetensors 6 MBpatches/wsp_s9_error_recovery_r16.safetensors 6 MBpatches/wsp_s1_think_r16.safetensors 6 MBpatches/wsp_s2_restraint_r16.safetensors 6 MBpatches/wsp_s3_router_r16.safetensors 6 MBpatches/wsp_s6_handoff_r16.safetensors 6 MBpatches/wsp_s8_grounding_r16.safetensors 6 MBpatches/wsp_s4_mid_code_r16.safetensors 5 MB有文件?提交种子。
提交种子google/gemma-4-12B-it