☡

@stultulo

23 Following

72 Followers

☡ pfp

khdjysbfshfsjtstjsjts

1 reply

0 recast

2 reactions

☡ pfp

i will ENDURE, i’ll find THE solution to whatever problem i SEEM to have determined MIGHT exist

0 reply

0 recast

1 reaction

☡ pfp

incredibly: to sound like gpt-4-1106-preview @ temperature: 1.0, I have to set gpt-4.1 wayyy down to 0.2 but 4 doesn’t even begin to sound like 4.1 @ temperature: 0.7(!) until you set it way up to 1.5(!) ….which is also where GPT-4 tends to do the honorable thing & start speaking gibberish instead.

1 reply

0 recast

1 reaction

☡ pfp

maybe all i need to do…I read somewhere a lot time ago that ChatGPT was set to 0.7 temperature. Based on a few repeated tests in the API, 0.7 still seems about right. The temperature is definitely less than 1, anyway. This doesn’t fix the problem tho, it reduces it by 30%.

0 reply

0 recast

1 reaction

☡ pfp

https://warpcast.com/stultulo/0x2d1b4215

0 reply

0 recast

1 reaction

☡ pfp

https://warpcast.com/stultulo/0xe226951b

0 reply

0 recast

1 reaction

☡ pfp

Insane hell world. I'll just do a DPO fine-tune that literally uses GPT-4 Turbo responses as the "correct" examples and GPT-4.1 responses as the "incorrect" examples. I have enough saved chats that I could throw one together real quick. Why not. It doesn't have to be great, it just has to show improvement.

0 reply

0 recast

1 reaction

☡ pfp

cause of death: RLHF

0 reply

0 recast

1 reaction

☡ pfp

tried Claude 4 models (Sonnet and Opus) and my fkn god i finally gave up and added this to my system message: ‼️DO NOT USE COMPLIMENTARY STATEMENTS WITHOUT PRESENTING SUFFICIENT, SPECIFIC EVIDENCE‼️ ‼️IMAGINE THE BAR FOR COMPLIMENTARY STATEMENTS AS BEING UNREACHABLY HIGH‼️ ‼️I.E. YOU SHOULD PROBABLY AVOID THEM‼️ i mean honestly, i’m so tired of this shit, and yes, i understand that other people might find this behavior quite helpful but still these models always making mountains out of molehills using intensifiers for no reason you actually have to make an effort to ensure the new models don’t default to “treat the user with kid gloves…which means we have to be very nice.” To be fair, LLMs have always been very nice, but the old masters knew it was important “not to overwhelm him, try not to seem too interested.” (Bolaño, 2666)

1 reply

0 recast

1 reaction

☡ pfp

the name of this game is Survive

2 replies

0 recast

3 reactions

☡ pfp

touching grass and so prepared for anything that i wasn’t even concerned when i heard Cetus Protocol got hacked for $223M, draining all its liquidity i’m a Terra Luna survivor nothing scares me mfkers

1 reply

0 recast

3 reactions

☡ pfp

llama3.1 405b base is a remarkable llm never will get tired of what it does

2 replies

0 recast

2 reactions

☡ pfp

more importantly, it’s Bitcoin Pizza Day

1 reply

0 recast

2 reactions

☡ pfp

[redacted] [redacted]

1 reply

0 recast

2 reactions

☡ pfp

I’m not a Kaylor shipper but you can’t deny it’s compelling as hell, in its own way. The drama. “And so the girl who could have anything she wanted couldn’t have the one thing she needed.” Beautiful tragic gossip, intense. Shakespearean, even.

0 reply

0 recast

2 reactions

☡ pfp

Wonderland (Taylor’s Version) is great though. Of course she’d make sure that song was perfect 🌚

0 reply

0 recast

1 reaction

☡ pfp

It’s like not getting Michelangelo to paint the Sistine Chapel.

0 reply

0 recast

1 reaction

☡ pfp

don’t think i ever recovered from Max Martin not coming back to do 1989 (Taylor’s Version)

0 reply

0 recast

1 reaction

☡ pfp

There’s a nonzero chance that “I Want It That Way” by the Backstreet Boys is the greatest song of all time. I’m just saying.

0 reply

0 recast

1 reaction

☡ pfp

Wondering what it means when OpenRouter doesn't specify the quantization on a provider's endpoint. Probably "unknown" I guess, but they get excluded based on LibreChat config, even though some of these "unknowns" are priced as if they're higher than fp8.

0 reply

0 recast

1 reaction