3 ms·
> A simple linear combination of every weight did not degrade the performance of the model, but enhanced it. Enhanced it on a couple benchmarks, supposedly. T
by Aurornis 5mo ago
> A simple linear combination of every weight did not degrade the performance of the model, but enhanced it.
Enhanced it on a couple benchmarks, supposedly.
The game is to turn knobs until you get a benchmark run that shows an improvement, then ship it. There are a lot of fine tunes and chimera models on HuggingFace that are supposedly better at some specific test, but when you use them for anything else they're usually worse.
This happens with a lot of the models that are modified to remove censorship. They succeed in getting the model to emit previously censored outputs, but the overall output quality decreases.
- andai 5mo agoThey seem to have deleted most of the README now, but the archived version has benchmarks. https://web.archive.org/web/20260614082641/https://huggingface.co/prefeitura-rio/Rio-3.5-Open-397B https://web.archive.org/web/20260614082641/https://huggingfa... And the Nex benchmarks for comparison https://huggingface.co/nex-agi/Nex-N2-Pro https://huggingface.co/nex-agi/Nex-N2-Pro Rio seems to be about halfway between Qwen 3.5 and Nex, as you'd expect?
- manquer 5mo ago> game is to turn knobs until you get a benchmark run that shows an improvement, then ship it i.e reinforcement learning against a weak reward function - benchmark is insufficiently complex and is not representative of the real world sufficiently. The "game", i.e. decision tree can be modeled as a multi-arm bandit problem, to deploy finite resources ( compute) toward exploitation/exploration . The main issue is each training / fine-tune is very expensive so number of chances at the slot so to speak is pretty limited today.
- monster_truck 5mo agoI don't think your last point is correct. Ablation, when done correctly, seems to increase the quality and typically also the performance too.
- Aurornis 5mo agoAbliterarion is a brute force technique that removes or silences parts of the model. It reduces performance because the abliterated elements aren’t perfectly isolated to censorship so other aspects suffer. Many of the “uncensored” model providers also do some fine tuning on the models. Some of them target better benchmarks or other measures, but outside of the benchmarks and metrics they’re fine tuned for they are generally noticeably worse than the original model.
- yowlingcat 5mo agoThe kind of abliteration you are mentioning is no longer state of the art or the most common form of removing the refusal layer in most models. Your your understanding was up to date about a year and a half ago, but has been out of date since after that.
- ls612 5mo agoNowadays it is that Heretic tool is it not? I’ve seen Gemma models uncensored with it.
- weitendorf 5mo agoUnrelated but I’ve been putting off learning about post-abliteration technique and want to use it for an upcoming open source “retraining” project I have on my backlog. I’m not interested in the refusal layers though, more like deep fine tuning but in a way that might let me prune out or consolidate layers, if that makes sense? Do you have any pointers or links to the current SOTA in this area? I guess I’m looking for a kind of bulk/sticky dropout (which was in fashion way back when I studied DNN in school).
- avadodin 5mo agoWhat OP is describing wasn't called abliteration at all. Abliteration whilst a neologism implies a surgical ablation of refusal. Earlier approaches post–trained the model to refuse less and, much like other kinds of fine–tuning, it degraded performance. They were "uncensored". Abliteration has seen some improvement to this day but it always was close to equivalent performance to the original when compared to those earlier techniques.