Vizuara AI Labs · closed-book QA fine-tune

SLM‑125M QA

Our 125M legal base model, fine-tuned to answer questions closed-book: no context passage, it answers from its own weights and refuses when it does not know. Trained on 7,141 self-contained, faithfulness-judged, decontaminated Q&A pairs (10% refusals). It is tiny, so expect fluency over reliable recall.

125M
parameters
16,384
vocab
6.06
QA val ppl
7,141
QA pairs
closed
book
Validation metrics along the 125M lineage
Each perplexity is measured on that stage's own validation set, so read the trend as 'how well the model fits its own stage's data', not as one curve on one dataset. DPO and RLAIF optimize preferences rather than likelihood, so they log preference margin and reward instead of perplexity. Click a stage to open that model.
Base
ppl 8.36
pretrain val
QA SFT
ppl 6.06
QA val
Instruct
ppl 6.85
instruction val
DPO
margin 75.4%
preference val, no ppl
/
RLAIF
reward 9.9→11.7
RM reward, no ppl
RAFT on DPO
ppl 2.01
RAFT val
/
RAFT on RLAIF
ppl 2.04
RAFT val
ask a legal / financial question
ready
The answer will appear here.

What this is closed-book QA

A supervised fine-tune (SFT) of the 125M base on closed-book Q&A. Unlike a grounded/RAG model, no passage is supplied at inference: the model answers only from what it internalised during pretraining. At ~125M parameters (~32MB of usable knowledge) it speaks the legal register well but does not reliably recall specific facts, which is exactly the limitation the larger models and the alignment stages (DPO, RLAIF) explore.

Served scale-to-zero on Modal, so the first question may take ~20–30s while the model wakes.

what this model cost to build
$121.48 total to assemble, on real Modal billing
Data & tokenization $2.15Pretraining $118.64Fine-tuning (SFT) $0.69
StageCostShare
Data pipelinecorpus clean + dedup + 16K tokenizer + tokenize (CPU)$2.151.8%
Pretraining10 epochs on 8x H100, 21.9B tokens seen, incl. smoke runs$118.6497.7%
QA dataset (shared)1/9 share of the 7,141-pair closed-book QA set$0.400.3%
Fine-tuning (QA SFT)full fine-tune on H100$0.290.2%
Total to assemble$121.48100%

Pretraining, with the data pipeline that fed it, is 99.4% of the bill. Everything after it (fine-tuning) came to $0.69. Teaching a model to behave is cheap; teaching it to know is not.

Figures are actual invoiced Modal usage (GPU + CPU + memory) for this model’s lineage, not estimates. Shared inputs (the QA and preference datasets) are charged at this model’s share, and their generation also used ~$1–2 of external LLM API calls, included above. Serving is billed separately and scales to zero.