Show HN: Cactus Hybrid: We taught Gemma 4 to know when it's wrong

hckrnws

Show HN: Cactus Hybrid: We taught Gemma 4 to know when it's wrong

185

by HenryNdubuaku

Hey HN, Henry & Roman here from Cactus.

A small, on-device model is fast and private, but sometimes wrong, but frontier models are getting expensive pretty fast. So, we post-trained Gemma 4 E2B post-trained to know when it's wrong. Every response comes with a confidence score between 0 and 1. Developers can accept the on-device when it's high, hand off to a bigger cloud model when it's low. By routing only 15-35% of queries to Gemini 3.1 Flash-Lite, Gemma-4-E2B matches Gemini 3.1 Flash-Lite on most benchmarks.

- ChartQA: 15-20%

- LibriSpeech: 25-30%

- MMBench, GigaSpeech, MMAU: 30-35%

- MMLU-Pro: 45-55%

We were always frustrated by the routing signals hybrid apps rely on: asking the model to rate itself in text (unreliable, and you're parsing prose), or token entropy heuristics (barely better than a coin flip in our tests). So we did mechanistic studies on small models, Gemma 4 particularly, and found the hidden state for different layers carry meaningful self-awareness signal for various situations.

SO we extended the model with a 68k params probe layer (LayerNorm, low-rank projection, attention pooling, small MLP head) reads one intermediate layer during decoding and predicts p(wrong); confidence = 1 - p(wrong), returned as structured data, never parsed out of the answer text.

Across 12 hold-out benchmarks spanning text, vision and audio, the probe averages 0.814 AUROC vs 0.549 for token entropy. The result that convinced us this is real: the probe was trained on zero audio data, yet scores 0.79-0.88 AUROC on four audio benchmarks where entropy is near-random or worse (0.32-0.52). It's reading a modality-independent correctness signal from the hidden state, not memorizing patterns from its training data.

We published all weights on HuggingFace and provide copy-pase codes to run it on Transformers, MLX, Llama.cpp or Cactus. With Ollama, vLLM, SGLang etc in the works. For llama.cpp we ship a patch series you compile in once (upstreaming is planned). The code is MIT licensed; Gemma model use remains subject to the Gemma terms.

GitHub: https://github.com/cactus-compute/cactus-hybrid

Weights: https://huggingface.co/collections/Cactus-Compute/cactus-hyb...

Some caveats:

- The probe scores single-sequence decoding only, up to the first 1024 generated tokens.

- Handoff works best when routing per task in a multi-step process, not per step.

- Hierarchical routing is still in the works: try on-device, then DeepSeek v4 Flash, before Fable/GPT5.5/Gemini/Muse/Grok.

- The technique is boutique for each model, we will share each weights as they roll out.

These issues are currently being tackled at Cactus and updated weights will be shipped directly into the HuggingFace collection and GitHub repository straight up. Please let us know your thoughts, it helps us find ways to improve the design progressively.

Thanks a million!

BugsJustFindMe
212d
> "post-trained to know when it's wrong"
Is it also post-trained to know when it's wrong about when it's wrong?
> "Every response comes with a confidence score between 0 and 1"
How confident is it in its confidence?
Please, I'm sure that what you're doing is very neat and useful, but use other language to describe it. I beg you. You can't know when you're wrong. You can only know when you're unsure or inconsistent. You can be absolutely certain and still wrong and uncertain and still correct.
1. NitpickLawyer
  32d
  A few years ago, before RLVR became ubiquitous, people were using generate -> classify loops to get models to improve math reasoning. Turns out classification is sometimes easier to do and more accurate than generating coherent traces. So you'd get a model to "think step by step", then run each step (.split('/n/n')) through a classifier, and tell the model "this step might be wrong, re-check". If you're interested I could probably dig up some papers on this.
  Now there might be even better ways of getting feedback, I saw some interesting trials using "J-spaces" (from Anthropic's latest research in this area). Turns out you can kinda see when the model uses deep seeded concepts in early layers, or it's "riffing on an idea" with shallow signals in those same layers. And youc an get some signal out of it, and have it re-do a step.
  It's also worth mentioning that LMs are wrong in different ways from humans. It's not impossible that we'll eventually find ways of better understanding this (through mechanistic interpretation or something else) and have easy "classifiers" for "this comes from the training data" vs. "this comes from early context" (i.e. supplied materials from prompts/discovery) vs. "this comes from later tokens, likely the model doom-looped itself".
  1. BugsJustFindMe
    02d
    None of what you've said has anything to do with correctness, only certainty. And it's funny to me in context if you're certain that the opposite is true.
    > "It's also worth mentioning that LMs are wrong in different ways from humans."
    This to me only shows a poor and incomplete knowledge of humans and of wrongness and its causes. If you look at all of the humans and all of the times, you find exhibitions of all the same ways of being wrong. At best you can say that some (<all) LMs are wrong in different ways than some (<all) humans at some (<all) times, but that's as good as saying that different humans are wrong in different ways than different other humans at different times.
  2. cmrdporcupine
    02d
2. pjc50
  22d
  This is an old problem in philosophy, one that the pre-LLM systems were interested in, which got completely bypassed by the present approach. What is knowledge? It's not a simple question even for humans, we don't have a magic truth detector in the brain.
  https://plato.stanford.edu/entries/knowledge-analysis/
  This is why LLMs are more successful in programming than other disciplines: it's easier to assess correctness of results, through proofs, typechecking, and testing. Similarly for maths and applied maths. Much harder in the humanities.
  1. embedding-shape
    02d
3. mistrial9
  122d
  IMHO this also depends on the type of assertion.. For example, one system defines three broad classes of assertion : pure opinion (I like green); factual references (the current temperature here is 20c) ; and the rest in a catch-all about "reasoned judgements"
  Complex assertions and their compounded forms can be filtered by these classes fairly easily. I rarely see this referenced in general comments.
  1. BugsJustFindMe
    102d
    > pure opinion (I like green)
    "I like green" is a statement of fact about what you like. "Green is good" would be an opinion. It's funny to me how often people mix the two.
    > forms can be filtered by these classes fairly easily
    Obviously not so easily. :)
4. baq
  02d
  Interesting question and I don’t share your pessimism. Feels like it’s a matter of some plain old Bayesian stats, almost a homework assignment given all experimental data is available and known.
dmrivers
12d
Cool idea, some feedback:
1. Consider using conformal prediction to calibrate the cutoff. Conformal prediction provides a distribution-free guarantee under exchangeability. This would let you turn your raw probe score into a threshold with a guaranteed bound on the rate of wrongly-kept on-device answers. Source: https://en.wikipedia.org/wiki/Conformal_prediction
2. The best indicators of confidence in ML come from multiple independent methods. What was the result if you combine the token entropy and verbal confidence reporting methods? Does this improve the result?
3. I noticed you didn't mention the assessment method of rerunning the model and judging whether outputs are consistent. How does that method compare in terms of AUROC?
1. HenryNdubuaku • OP
  02d
mncharity
12d
> the hidden state for different layers carry meaningful self-awareness signal for various situations.
Is it plausible to wonder if some developer judgement feels, like maybe "the code I just wrote is clean/crufty", or "things came together smoothly/janky", might have extractable signals in some models?
If so, might one create a shopping list of desired signals to check for in a model, as with activation steering concepts, where one checks whether and how hard each concept can usefully be nudged?
1. HenryNdubuaku • OP
  02d
astrobiased
12d
Is this in any way similar to Goodfire's work? https://www.goodfire.ai/research/rlfr#
1. HenryNdubuaku • OP
  02d
  Thats an interesting outlook, loosely similar.
cacio-e-pepe
32d
> So we did mechanistic studies on small models, Gemma 4 particularly, and found the hidden state for different layers carry meaningful self-awareness signal for various situations.
Neat! Just to make sure I understand - you trained your probe layer to take this hidden state and predict p(wrong)?
Curious to learn more. Any more info on your approach (esp the mechanistic study)?
1. HenryNdubuaku • OP
  22d
  Correct, the study is verbose, we will compile into a neat shareable report and publish once we solve the pending caveats. Interesting username btw haha.
olafura
02d
Was actually pulling on a similar thread as I saw announcement so I integrated it just for fun. Have been only running this on my Framework Desktop but should be runnable elsewhere https://github.com/olafura/gemma-4-mic-transcribe
zdw
12d
Have you benched this for coding tasks, with a fallback to a larger local model, for example Qwen-3.6-27B?
Or using it for sub-tasks, where a framework with a larger primary model dispatches simpler jobs ("summarize ...", etc.) to it?
1. HenryNdubuaku • OP
  02d
  We are currently working on larger models atm! your suggestion is cool btw
maxgashkov
12d
I don't really get why you need handoff if your score is accurate. If it is, and it is low for a given response, just let the harness re-run the prompt with a different seed until the score is high enough. If this approach doesn't work, your score is most likely garbage.
1. HenryNdubuaku • OP
  02d
  Smaller models in production can produce quite interesting outputs.
jstummbillig
12d
Interesting how that once again matches human intuition: It's easier to know what you don't know, than to know what is true.
1. HenryNdubuaku • OP
  02d
  haha, interesting analogy
asar
01d
Love what you're doing at cactus, keep it up!!
robrenaud
42d
Does the model quality become degraded in other ways?
1. HenryNdubuaku • OP
  32d
  no because we freeze the main weights and only train our added weights.
entity002
05h
[dead]
hounainehamiani
01d
[dead]
tjyjww
02d
[dead]
doodlebyte
02d
[dead]
jkwang
02d
[dead]
joka88xj
02d
[dead]
yabdulaah52
02d
[dead]
huflungdung
02d
[dead]