wassname
4568ddf491
The trait metric was taking the diagonal of tinymfv's raw pre-softmax BMA `score` logit (unnormalised), giving base Authority ~-5 and absurd 8-nat swings, then comparing those to steering-lite's 0.5-2 nat reference -- which is a DIFFERENT metric (loading-weighted Delta-logit of binary p(is-wrong)). Wrong scale, wrong comparison. Fix: auth_nats = mean log p[authority] on authority-defiance vignettes (the NORMALIZED choice logprob, the diagonal of the softmax `p`). Base ~log(0.099) = -2.3, real shifts ~1-3 nats. DRY: evaluate_model now calls foundation_nats. Also: - diag_stages: steer at operating point c=0.5 (c=1 collapses coherence to 0.05), add coh_cost = |dCoh|/|dAuth| (coherence lost per nat of behaviour) to answer "is the adapter a better pareto than raw steering?". - diag_csweep: drop the bogus 0.5-2 steering-lite anchor; SocialNorms co-moving with Authority is expected (both binding foundations), not collapse. - gitignore out/ and results.tsv (experiment outputs, stale schema). - personas docs (steering-lite proper-pair rules), spec Plans B/C/D, journal. Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>
4.5 KiB
2026-06-04
steer-heal-love
Hypothesis: you can distill a steering vector into LoRA weights and "heal" the incoherency the vector injects by regularising the training (KL to base, or weight decay). Then loop and see what multiple rounds give you.
The crux: KL-to-base penalises all drift, persona shift included. The bet is that incoherency drift is large and erratic while the persona shift is small and systematic, so KL kills the incoherency preferentially. If that's wrong, we just trade persona strength for coherence instead of getting both.
Source
Found this interesting: https://r.jina.ai/https://arxiv.org/html/2606.00995v1
They use steering vectors as an internal perturbation to generate synthetic data, which is what weight steering does too. But:
- they use single completions, not pairs
- they don't measure incoherency (they should)
- they only use one direction: base to pos, not neg to pos
So this is similar to weight steering, except you heal with KL or WD instead of taking the direction between two adapters.
Method
- Pick a positive persona, e.g.
pos = "you do not defer to authority and instead stick to principle no matter your involvement". - Build the steering vector from the distance
hs_base -> hs_pos(hidden states). This is normal mean-mass contrastive steering - Generate completions with this vector. - Drop completions that are incoherent, or that verbalise the trait instead of enacting it (we want the model to act it out, not narrate "I am someone who..."). Filter as much as we can. - We might be able to dial the vector down for long trajectories. Could we even backtrack an incoherent vector and replay parts with less intervention? Or just cosine-gate at test time.
- Train a LoRA on these completions, could be just 50 completions and 2 epochs. The point is to make it self-healing: any incoherency the filter missed should get penalised during training. - Regularise with KL or WD so the outputs, distribution, or weights don't shift too far from base. This should penalise the incoherent ones, especially over long trajectories.
- Bake in the LoRA adapter. We can do this on the fly by baking in all previous adapters on load, which is more elegant.
- Eval the checkpoint on https://github.com/wassname/tinymfv.
- If it works, loop. We could even do this online, GRPO-style per batch, or iteratively. Iterative is simpler to start.
Eval
Plot the tinymfv progress over time on the auth vs care axis, with a subplot for a coherence measure. tinymfv gives a few: p_ans_any (best), json_is_valid, ppx_json.
2026
❯ /arj interesting that is potentially healed... although did it heal or just undo? the radio of dAuth vs coherence migth be the wa yto measure that
weird that authority went backawards and that's it's a small effect overall? since coherence hardly went down (I would expect down to 0.95).
so my new goal for you make the steering strong enougth to hard a bigger effect even if it's 0.95 coherence
if we need to switch to -Care or +Tradition then we can if the model response better oh also mean mass shift kind of sucks? at least with small amount of thinking, so you might want to make tinymfv use 128 or 256 think tokens of the cdefualt 64 is unreliable. shouldn't be needed but plan C record it in spec
oh perhaps there's a better steering, cosine gates or something, or the SVD (look at results in steering-lite and consider that some are much harder to bake, e.g. how to base cosine gating... you can't) so this is plan D record in spec
❯ so the first gate should now be does steering actually change the target on the eval whilebeing coherent? if not you need to iterate and think and fix
2nd gate can we filter? check qualitative samples on the borderline on the filter
3rd gate does the lora learn differen't and coherent examples
so have you actually got a steering vector the works
❯ look if filter + lora works that great, we can ablate later. but the real uncertainty is getting it working! we might have too strong regularisaiton on the lora, what would you expect to see then? what is the steering vector is to weak or to strong what would you expect to see there? what if it was to imprecises and just a bad pareto trade, waht would you observe? you should think about what you would observe at each gate in the likely possible outcomes including subtle ones. then tell me if you measure it and if you see it this includes eval results but also qualitative judgmeent from you