jun

just doing things and looking at data

[jun @junkliao]

@dvruette @Creative_Math_ twitter rederives "doing research" from first principle

[jun @junkliao]

@FakePsyho is there a particular reason they are all japanese?

[jun @junkliao]

@mmaaz_98 @regulargio today i taught my agent about smoke breaks and he is now french

[jun @junkliao]

@jackson_stokes can you explain the loss function and it's just the reverse kl

[jun @junkliao]

@notmoeezm @paularambles chat gpt is def a cog-like scaling of mechanical process, i would not underestimate scaling mechanical processes

[jun @junkliao]

@KoSan15536261 it’s 5

[jun @junkliao]

@beauti_animals are you a genius and it’s straight up just mctavish

[jun @junkliao]

@Leophilius not a single grown up believes in the p or ~p lie

[jun @junkliao]

@_sholtodouglas when i hit the rate limit after 2 prompts

[jun @junkliao]

any papers with main results benchmarking math on qwen models i don’t trust, these models are so overfitted on the test sets that you could do any training and it would improve performances

[jun @junkliao]

@notmoeezm @nickaturley 👀

[jun @junkliao]

@usr_bin_roygbiv @scaling01 no, have you?

[jun @junkliao]

@slitkrisXD tehhh

[jun @junkliao]

just got out of the verl dependency hell i see no god up here other than me

[jun @junkliao]

@TheMingjie how many turing awards does your school have?

[jun @junkliao]

give me 7 days to build an app and i will spend the 6 first days refining my prompts

[jun @junkliao]

@KoSan15536261 mfs after one loss

[jun @junkliao]

@KoSan15536261 mfs after one win

[banteg @banteg]

anthropic running the exact same marketing playbook with every release. “our model is so capable and dangerous, ahh we are afraid to release it”. just put the model in the bag lil bro.

[jun @junkliao]

@notmoeezm agreed