TP
@jensensjacket.bsky.social
Workin on RL (HF/AIF). Interested in AI Risk Management and Mech Interp. Not Intelligent.
I suppose it matters if agents have sufficiently different capabilities/authorization. Or instructions (help with eval). But I don’t think I’ve seen any real test at this level of granularity.
November 23, 2024 at 12:04 AM
I suppose it matters if agents have sufficiently different capabilities/authorization. Or instructions (help with eval). But I don’t think I’ve seen any real test at this level of granularity.
update: my kids hate it
November 22, 2024 at 7:13 PM
update: my kids hate it
great place to integrate with Illuminate
November 22, 2024 at 5:16 PM
great place to integrate with Illuminate
👋
the starter packs are key
the starter packs are key
November 22, 2024 at 4:12 PM
👋
the starter packs are key
the starter packs are key
Miserable. I need to continue building an ARC challenge captcha
November 22, 2024 at 4:11 PM
Miserable. I need to continue building an ARC challenge captcha
😂 What labels will you advertise?
November 22, 2024 at 4:08 PM
😂 What labels will you advertise?
Do you see any problems with coherence as vocab collapses?
November 22, 2024 at 4:00 PM
Do you see any problems with coherence as vocab collapses?