Nagi-Big

System One typed-decision model (choice / score / noul) β€” Qwen3.5-4B + PiSSA LoRA + Choice-B option scoring. Soft-distilled from teacher ensemble (DeepSeek flash + GPT-5.6-Luna). No text generation; calibrated option probabilities.

Also known as: Track-G / hail-g1 / nagi-t4-qwen-pissa.

Load

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

tok = AutoTokenizer.from_pretrained("Qwen/Qwen3.5-4B")
base = AutoModelForCausalLM.from_pretrained("Qwen/Qwen3.5-4B", dtype="bfloat16")
model = PeftModel.from_pretrained(base, "nagisanzeninz/nagi-big-v0")
# optional: model = model.merge_and_unload()

Inference recipe: nagi.render.render_nagi_prompt (definition + 2 demos + state + options) β†’ last-token logits over letter ids A,B,C… β†’ softmax over K options. See scripts/t4_qwen_pissa.py and scripts/t4_leave_task_out.py in nagisanzenin/nagi.

gold_ltout LTO (2700, 6 public unseen tasks)

Task n hard_acc
Amazon 500 0.948
DBpedia 500 0.900
CoLA 500 0.724
MRPC 500 0.704
20NG 400 0.630
QQP 300 0.350
macro 2700 0.709

3-way (same gold_ltout)

Nagi-Big Laya Jev
LTO hard 0.709 ~0.35 0.773
p50 latency 84 ms (L4 eval) 33 ms 501 ms (API)
marginal cost self-host self-host API

Beats Laya by ~+36 pts on LTO. Within 6.4 pts of Jev. Gates: USEFUL (β‰₯0.60) βœ… Β· NEAR (β‰₯0.70) βœ… Β· WIN (β‰₯0.75) ❌.

Train

  • 25,000 items from train_t4_final (1,443 schemas: public multi-task + hail-C novel schemas + burst/synth)
  • PiSSA r=32 Ξ±=64 pissa_niter_4 dropout=0, lr 1e-4, bs 2, 1 epoch, max-len 768
  • Loss: soft-label KL L = βˆ’Ξ£ q log p (no hard warmup)
  • H100 BF16 Β· wall 4816s

Sibling

  • Nagi-Smol β€” 421M M2β€² dual encoder, ~11 ms, LTO 0.462
  • nagi-t4-m2p-v0 β€” prior T4 M2β€², LTO 0.481
Downloads last month
-
Inference Providers NEW
This model isn't deployed by any Inference Provider. πŸ™‹ Ask for provider support

Model tree for nagisanzeninz/nagi-big-v0

Finetuned
Qwen/Qwen3.5-4B
Adapter
(1,216)
this model