Rendered at 12:03:52 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
phantompeace 3 hours ago [-]
Considering the very small difference between just SFT on the student model as compared to SFT + DPO on a proxy, doesn't it make sense to concentrate on ensuring the SFT dataset is perfect rather than sorry about DPO etc? And just train directly on the student model?
dmezzetti 11 hours ago [-]
Well-Read Students Learn Better: On the Importance of Pre-training Compact Models
Why is this published again? Is this a reference to recent events?
babelfish 12 hours ago [-]
I just saw some post about it on Threads and found it interesting so decided to share!
tough 8 hours ago [-]
My best guess is this is a reference to the recent accusations from Anthropic of chinese labs ¨distilling¨ on their models
swingboy 2 hours ago [-]
And it’s a paper from Alibaba researchers, the company/lab that Anthropic called out by name.
10 hours ago [-]
StreamCtx 1 hours ago [-]
“Relevant to anyone building failure-attribution systems for agent pipelines — black-box distillation techniques here could feed into causal attribution models without needing white-box access to the underlying model.”
duendefm 13 hours ago [-]
The Chinese are really going strong on destroying the American AI economy bubble. Honestly, despite the fact that I'm totally pro USA and anti China, I think we should help them crashing the American AI bubble. They are controlling everything and we can't even buy a new computer nowadays while getting no benefit from this. I wish some influential programmers stimulated coders everywhere to skip Claude and Chatgpt subscriptions for Chinese ones, at scale. If we programmers united we could help this bubble burst, I'm sure.
laichzeit0 2 hours ago [-]
The US government will do the job of destroying the American AI economy through their export controls.
anax32 4 hours ago [-]
The US "product machine" is so strong. They really know how to do frictionless signup and vendor lock-in on the corporate side.
nozzlegear 13 hours ago [-]
> skip Claude and Chatgpt subscriptions for Chinese ones, at scale. If we programmers united we could help this bubble burst, I'm sure.
I'm doing my part!
cynicalsecurity 13 hours ago [-]
[flagged]
anon373839 12 hours ago [-]
Dario, is that you? Is Anthropic’s next ploy to seek support via the culture wars?
girvo 9 hours ago [-]
Why would I care about Christian morals? In fact from what I can see of the US, you don’t have them either.
duendefm 13 hours ago [-]
Nvidia, Anthropic and OpenAI are controlling everything, and nothing is improving for everyone, quite the opposite. So I just hope they crash to the ground.
13 hours ago [-]
gmerc 13 hours ago [-]
lol Christian Morals. Epstein and his best buddy running the show tells you all about this
big-and-small 12 hours ago [-]
What Epstein and buddies were doing was very... Christian...
Virgin Mary was very young in the events you know.
LNSY 12 hours ago [-]
"They don't have Christian morals" -- does that mean they don't commit genocide and fuck kids? Because that sounds like a point for them
linolevan 13 hours ago [-]
Can we note that this is a 2024 paper in the title?
Related paper that's a good read: https://arxiv.org/abs/1908.08962
I'm doing my part!
Virgin Mary was very young in the events you know.