Skip to content

Merge research-pipeline: better router evaluation pipeline and readme.md - #8

Merged
yl231 merged 14 commits into
mainfrom
research-pipeline
Nov 4, 2025
Merged

Merge research-pipeline: better router evaluation pipeline and readme.md #8
yl231 merged 14 commits into
mainfrom
research-pipeline

Conversation

@yl231

@yl231 yl231 commented Nov 4, 2025

Copy link
Copy Markdown
Contributor
  1. Updated the README.md for easier usage.
  2. The pipeline is now designed for better evaluation of incoming new routers.

@yl231
yl231 merged commit 2166456 into main Nov 4, 2025
10 checks passed
@yl231
yl231 requested a review from jiarong0907 November 4, 2025 03:04
@jiarong0907
jiarong0907 deleted the research-pipeline branch November 10, 2025 17:25
ypollak2 added a commit to ypollak2/RouterArena that referenced this pull request Jun 2, 2026
Iteration v2 after analysing routing patterns vs Sqwish (RouteWorks#1).

Pool changes:
  - Added qwen/qwen3-next-80b-a3b-instruct (Sqwish's pick on
    SuperGLUE-ClozeTest and a subset of LiveCodeBench)
  - Removed anthropic/claude-sonnet-4 (low ROI: 8% of routes
    at 8x cost for <1% accuracy lift at full scale)

Routing changes:
  - SuperGLUE-ClozeTest → qwen3-next-80b (matches Sqwish: 33/36 wins)
  - QANTA → deepseek-v4-flash (was claude; pool ceiling stays
    the same for this dataset, claude wasn't earning the premium)

Result: Arena 0.7046 → 0.7081 (+0.0035), cost $0.31/1K → $0.149/1K (52%).
Accuracy: 71.67% → 71.20% (-0.47), but cost reduction nets gain
in Arena Score.

Position: ~RouteWorks#8 (above Auto Router 0.7005, below R2-Router 0.7160).

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
yl231 added a commit that referenced this pull request Jun 13, 2026
llm-router by @ypollak2 — open-source MCP-server router (PyPI: llm-routing,
source: ypollak2/llm-router). Evaluated in #134 at RouterArena score 0.7126:

  Arena    71.26
  Accuracy 72.05
  Cost/1K  $0.20
  Opt.Sel  18.01
  Opt.Cost 20.46
  Opt.Acc  89.13
  Robust   30.00
  Abnormal 0

Slots in at #8 (between R2-Router 71.60 and Azure 70.42); ranks 8-20 each
shift down one, leaderboard grows to 21. Metrics from the final /evaluate
on the merged submission, verified byte-identical to main (no post-eval swap).

Co-authored-by: Louie Lu <yl231@datalab2.cs.rice.edu>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
@yl231 yl231 mentioned this pull request Aug 20, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants