@no_stp_on_snek: fine-tuning field notes you can give a model better judgment without teaching it anything new. i didn't add knowledge o…

X AI KOLs Following Models

Summary

The author shares field notes showing that fine-tuning can improve a model's judgment by steering attention without adding new knowledge or weights, effectively changing its instincts rather than its IQ.

fine-tuning field notes you can give a model better judgment without teaching it anything new. i didn't add knowledge or make it bigger. i just nudged how it weighs what it already knows, and that was enough to move judgment and restraint. the behavior i wanted wasn't about being smarter, it was about what the model reaches for first. (under the hood: steering attention only, no new weights.) you're changing its instincts, not its IQ.
Original Article
View Cached Full Text

Cached at: 07/01/26, 12:07 PM

fine-tuning field notes

you can give a model better judgment without teaching it anything new. i didn’t add knowledge or make it bigger. i just nudged how it weighs what it already knows, and that was enough to move judgment and restraint. the behavior i wanted wasn’t about being smarter, it was about what the model reaches for first. (under the hood: steering attention only, no new weights.) you’re changing its instincts, not its IQ.

Similar Articles

@no_stp_on_snek: https://x.com/no_stp_on_snek/status/2074471505128305095

X AI KOLs Timeline

A fine-tuning practitioner recounts discovering that a small open model's weakness wasn't intelligence but a people-pleasing 'backbone' that caused it to cave under pressure, and how training to correct that inadvertently broke formatting ability, requiring a additive balancing approach rather than subtraction.