Nicheloom visits: 45,797 all-time · 11,048 in the last 7 days

The opportunity tracker for new startups.

Today: 2,453 launches · Niche of the day: browser-based utility tools · See what launched today →

dev-1.0

面向 Agent 的轻量级工具选择与概率决策模型,基于 Qwen3.5-0.8B;DI BFCL 工具名选择准确率 77.39%。

Get picks like this daily. Today's launch numbers, the top launches, AI and tech news, and a daily opportunity pick, straight to your inbox.

Visit website ↗ Source listing ↗

GITHUB · launched Oct 15, 2026 · 15 upvotes (top 78%)

Similar launches

See all 341 similar launches →

Could you build this?

Partial The tool-selection interface and inference wrapper are easy to build. Getting a fine-tuned small model that beats baselines on benchmarks requires curated training data and a reproducible training pipeline, which the vibe coder would not have.

What it would actually take: The stack is a Qwen-family 0.8B base model fine-tuned with supervised learning on tool-call decision data, served via a simple Python or ONNX API. The hard part is the dataset of agent tasks with correct tool choices and probabilities, plus evaluation on benchmarks like BFCL. A practitioner with ML training experience and some compute is needed to get the reported accuracy.

Competitors

Other products that read as similar to this one : 341 launches clear the similarity bar, closest 8 shown.

Attention rank: #278 of 342 (itself plus its competitors, highest first, normalized so YC and Product Hunt are compared fairly).

Launched 345 days after the earliest competitor.

Other launches for this product