A3M Router submission - what is the official listing process?

[Das-rebel]

Question: How to get a router officially listed on RouterArena?

I've submitted PRs #113 and #120 to add A3M Router to RouterArena, but both were closed without merge.

My fork (Das-rebel/RouterArena) has the router working, but A3M is not appearing in the official leaderboard.

What is the correct process for official router submission?

A3M Router is an open-source parallel multi-LLM execution router with:

• Parallel execution across 47+ providers (vs sequential fallback in litellm/one-api)
• RouterArena robustness leaderboard [Setup] Pre commit #1 (76.43 score)
• OpenAI-compatible API, MIT license
• npm: npm i -g adaptive-memory-multi-model-router

I've also submitted PRs to:

• RouterBench (baseline submission)
• MMR-Bench (multimodal routing)
• LLMRouterBench (merged)

Should I be submitting benchmark results via a data file instead of a README PR? Please advise on the correct path.

Thanks!

[yl231]

Hi @Das-rebel — sorry for the delay on our end; that's on us, and I'm sorry it left #113 and #120 hanging until you closed them. Nothing was rejected — we'd be glad to get A3M evaluated and onto the leaderboard.

You actually had the process right: it's a prediction-file PR + /evaluate, not a README edit. Could you open one fresh PR to main with:

• router_inference/config/a3m-router.json
• router_inference/predictions/a3m-router.json (8,400 entries, generated_result populated)
• router_inference/predictions/a3m-router-robustness.json (420 entries)

Then comment /evaluate.

One ask so the number holds up: please submit a single clean run from one fixed router version (RouterArena is evaluation-only, so the router can't be tuned against the scores/labels), and re-run inference after any routing change so each prediction is paired with a generated_result from that same model.

Thanks for your patience!