cd /news/machine-learning/training-a-57m-edge-model-sft-tool-c… Β· home β€Ί topics β€Ί machine-learning β€Ί article
[ARTICLE Β· art-173192] src=dev.to β†— pub= topic=machine-learning verified=true sentiment=↑ positive

Training a 57M Edge Model: SFT Tool-Calling Done, Now Branching into 3 RLVS Loops

A developer trained Nova, a 57.2M-parameter edge model, through supervised fine-tuning with a syntactically verified tool-calling dataset and Program-of-Thought execution traces, reaching 100% deterministic tool-calling precision. Rather than continue single-track training, the developer cloned the base SFT checkpoint into three forks, each trained under a specialized Reinforcement Learning with Verifiable Supervision verifier covering cybersecurity, ARC-AGI, and balanced desktop execution, to avoid gradient interference between skills.

recommendation

Freeze a verified tool-calling SFT base and train separate RLVS specialist forks per skill instead of one mixed dataset, to prevent gradient interference in sub-100M models.

more in playbook: Fine-tuning and data
by read2 min views2 publishedNov 20, 2026

TL;DR: We just completed Supervised Fine-Tuning (SFT) on Nova, our home-grown 57M parameter edge model, hitting 100% deterministic tool-calling precision. Instead of continuing single-track training, we are cloning the base weights into three distinct forks, each trained under a specialized Reinforcement Learning with Verifiable Supervision (RLVS) verifier: Cybersecurity, ARC-AGI, and Balanced Desktop Execution.

When training a sub-100M parameter model (Nova is exactly 57.2M parameters), the biggest danger is format drift during structured JSON tool calls. Traditional LLMs often hallucinate brackets or invent parameter keys when parameters drop below 1B.

By curating a clean, syntactically-verified tool-calling dataset with Program-of-Thought (PoT) execution traces: With the base grammar and tool execution locked in, we reached a crossroads: How do you teach a 57M model high-level reasoning without bloating its parameter count?

Instead of forcing a single 57M model to master everything and diluting its weights, we are taking the identical base SFT checkpoint and cloning it into three parallel training tracks:

            β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”

              β”‚    Nova 57M Base SFT Checkpoint β”‚
                  β”‚   (100% Deterministic Tools)    β”‚

                  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

                                   β”‚ Clone Weights

    β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”

    β–Ό                              β–Ό                              β–Ό

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β” β”‚ Track 1: Nova-Sec β”‚ β”‚ Track 2: Nova-ARC β”‚ β”‚ Track 3: Nova-Balancedβ”‚ β”‚ RLVS Focus: β”‚ β”‚ RLVS Focus: β”‚ β”‚ RLVS Focus: β”‚ β”‚ β€’ Prompt injection β”‚ β”‚ β€’ ARC-AGI visual grid β”‚ β”‚ β€’ Low-latency Cowork β”‚ β”‚ β€’ Secure code audit β”‚ β”‚ β€’ Spatial transform β”‚ β”‚ β€’ General debugging β”‚ β”‚ β€’ Exploit verificationβ”‚ β”‚ β€’ Step-wise logic β”‚ β”‚ β€’ Real-time OS agent β”‚ β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜ β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

In typical supervised fine-tuning, training on diverse datasets causes gradient interference (learning security rules degrades spatial reasoning, and vice-versa).

By freezing the SFT tool-calling base and training three separate RLVS specialists:

For those training small-language models (SLMs) under 100M parameters: Drop your thoughts below!

── more in #machine-learning 4 stories Β· sorted by recency
── more on @nova 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain β€” perfect for shipping the agent you just read about.

$git push zahid main
β†’ Live at https://your-agent.zahid.host βœ“
Get free account β†’ Pricing
from €0/mo Β· no card required
LIVE [news/training-a-57m-edge-…] indexed:0 read:2min 2026-11-20 Β· β€”