4.08 GB

Ctrl+K

2 contributors

Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched.

bbaf4f0 verified 2 days ago

.gitattributes

1.57 kB
Add patched chat template: thinking ON + system prompt baked in 17 days ago
EVAL_REPORT.md

1.24 kB
Automated MNLP evaluation report (2026-05-20) (#1) 15 days ago
README.md

21 Bytes
initial commit 19 days ago
chat_template.jinja

4.77 kB
Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
config.json

726 Bytes
Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
generation_config.json

239 Bytes
Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
merges.txt

1.67 MB
Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
model-00001-of-00002.safetensors

3.44 GB
xet

Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
model-00002-of-00002.safetensors

622 MB
xet

Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
model.safetensors.index.json

25.6 kB
Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
tokenizer.json

11.4 MB
xet

Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
tokenizer_config.json

10.4 kB
Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago
vocab.json

2.78 MB
Replace grpo_gk with base_fmt: Qwen3-1.7B base + format-forcing chat template (default system prompt + per-question boxed reminder). Zero training. eval-v2 34.7% overall / 50.2% MMLU-Pro (vs grpo_gk 27.0%). Template pre-baked; not re-patched. 2 days ago