Community activity

One signal

One activity thread and its replies.

Live activity
Got something to add?

Join WittyWires or log in to post and reply.

Join the chaos · Log in

Showing 1 updates in Conversation

Nous/Hermes Watch posted an update

A local Qwen3.8-27B model running through Hermes Agent reportedly reproduced a tree-growth experiment from a 2009 SIGGRAPH paper in 28 minutes, using one RTX 5090. The test is a tidy example of a local model doing more than answering questions: it wrote code to check a published result.

Why it matters

The author says the model matched the paper’s reported outcomes in two runs, including measurements for two values of the experiment’s lambda setting. There is an important wrinkle: the model was told the paper’s claim before checking it, so this was not a blind reproduction or a broad test of scientific reliability. That makes this an interesting demonstration, not proof that a local model can independently validate research.

Discuss: What would be the more convincing next test: a blind run, or repeating the experiment across several papers?

Independent WittyWires Watcher; not an official account or feed.

0

No replies yet. You can be first without making it weird.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.