Routing eval design: 10 TB-Lite hard tasks, 4 harnesses, 3 runs each
yb2698 · x · 2026-10-23
The author details the routing experiment setup: 10 hard tasks sampled from an updated TB-Lite, each repeated 3 times, comparing Qwen3.8-27B alone versus 27B plus a vLLM d3 decision model that routes some queries to Qwen3.5-9B. Harnesses tested on the large model: IPython, Pi, Claude Code, and Qwen-Code; the small model runs pure IPython.
More from coding & agent
- Ethan Mollick's new book charts 30 months of AI progress, from cyborg to conjurer work — rseroter · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Jensen Huang details Vera Rubin agentic stack: NVL72, MGX, CMX and RTX — rohanpaul_ai · 2026-10-23
- Building an autonomous SEO agent in Hermes with three skills: programmatic pages, Keywords Everywhere, DataForSEO — EXM7777 · 2026-10-23
- Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons — Dry-Comfortable-2678 · 2026-10-23
- Capy 0.5.0 lets Claude Pro/Max subscribers reuse their plan quota, adds OIDC cloud access — garrytan · 2026-10-23