Symbolic Circuit Distillation: prove program to LLM circuit equivalence
Hi HN, I've been exploring various applications of formal methods to ML/interpretability and I've been hoping to get more eyes on the approach. I have been working on a small interpretability project I call Symbolic Circuit Distillation. The goal is to take a tiny neuron-level circuit (like the ones in OpenAI's "Sparse Circuits" work) and automatically recover a concise Python program that implements the same algorithm, along with a bounded formal proof that the two are equivalent on a finite token domain. Roughly, the pipeline is: 1. Start from a pruned circuit graph for a specific…
Symbolic Circuit Distillation: prove program to LLM circuit equivalence launched on January 6, 2026 with 16 votes, #777 of 14,584 launches that month and more than 90% of that year's launches.
In plain words
Symbolic Circuit Distillation is a tool for AI interpretability researchers that automatically converts neuron-level circuits from transformer models into equivalent Python programs with formal proofs of correctness. It works by extracting pruned circuit graphs from transformers, training surrogate networks to match their behavior, and symbolically verifying equivalence over finite token domains. The project is aimed at researchers exploring formal methods applications to machine learning and interpretability.
written from the facts on this page · September 2026
From the sources
Automatically extract executable programs from pruned mechanistic circuits, extending OpenAI's Sparse Circuits - neelsomani/symbolic-circuit-distillation
Mechanistic interpretability increasingly reveals small, crisp circuits that implement algorithmic behaviors inside transformers, such as delimiter matching, simple counting, or type tracking. However, converting these neuron- and channel-level circuits into faithful, human-readable explanations still requires extensive manual effort and bespoke reasoning. We introduce Symbolic Circuit Distillation, a method for automatically converting neuron-level circuit graphs into concise, executable programs with bounded formal correctness guarantees. Given a pruned circuit extracted from a sparse model or from dense-model feature spaces (e.g., SAE features or attribution graphs), we treat the circuit…from github.com
In the maker’s words, at launch
Hi HN, I've been exploring various applications of formal methods to ML/interpretability and I've been hoping to get more eyes on the approach. I have been working on a small interpretability project I call Symbolic Circuit Distillation. The goal is to take a tiny neuron-level circuit (like the ones in OpenAI's…
Does a similar job
all alternatives →
Duplicate 3 layers in a 24B LLM, logical deduction .22→.76. No trainingMar 2026 · github.com · ▲265I replicated David Ng's RYS method (https://dnhkng.github.io/posts/rys/) on consumer AMD GPUs (RX 7900 XT + RX 6950 XT) and found something I didn't expect. Transformers appear to have discrete "reasoning circuits" — contiguous blocks of 3-4 layers that act as indivisible cognitive units. Duplicate the right block and the model runs its reasoning pipeline twice. No weights change. No training. The model just thinks longer. The results on standard benchmarks (lm-evaluation-harness, n=50): Devstral-24B, layers 12-14 duplicated once: - BBH Logical Deduction: 0.22 → 0.76…
Distilling DeepSeek into GPT-OSS doesn't transfer censorship. Try itJul 2026 · ctgt.ai · ▲170We recently used DeepSeek V4 Flash as a teacher for finance tasks with GPT-OSS-120B. Distillation works well on this problem. At a constrained 8k token budget, our self-distilled 120B scores 83.61% on FinanceReasoning, above Kimi K3 (81.93%) and Inkling (65.13%). We released the 20B open weights. With V4 as the teacher though, we realized it would be timely to measure if the censorship characteristic of it transferred to the distilled version of the base model. tl;dr it didn't, the teacher answered politically sensitive questions 7 SDs differently than expected, but the distilled model's…
I made a little digital circuit simulator that operates on PNGs2022 · github.com · ▲259This is a little toy project of mine that lets you simulate digital logic graphs. It was inspired by Minecraft's Redstone and the Piet esolang. It's got some serious drawbacks-- you write circuits as PNGs and simulate them with a Python interface. It's slow to run and slow to experiment with. And it is certainly difficult to use for people with any kind of color blindness. But despite that, I hope this can still be a fun toy!
zkGolf – Competitive optimization of formally verified circuitsJul 2026 · zk.golf · ▲69Zero-Knowledge Proofs (ZKPs) let an untrusted proved show that computation was executed correctly without revealing the inputs to the verifier. However to prove anything, the computation first has to be expressed as a circuit: a system of polynomial equations (constraints) over a finite field. Circuits are the assembly language of zk and every constraint costs prover (and sometimes verifier) time, so production circuits are aggressively hand-optimized. Over the last months, we have been experimenting with writing formal specifications instead and letting LLMs produce the circuits: as long as…
DistilKitPlus, a distillation framework between any LLMs2025 · github.com · ▲10Over the past few months, I have built a distillation toolkit that supports cross-tokenizer distillation (e.g., distilling from LLaMA to Qwen vocab, or others). This approach has worked well on reasoning datasets like AIME, and we’ve validated on models like Phi and Qwen. We’ve also integrated Modal for quick deployment (with $30/month credits to try it out). Would love any feedback! GitHub: https://github.com/agokrani/distillKitPlus Docs: https://distillkitplus.mintlify.app/
ArcaneVM – A Fully Homomorphic Encryption Brainfuck Virtual Machine2019 · github.com · ▲131A Fully Homomorphic Encryption Brainfuck virtual machine - f-prime/arcanevm
More ai this month
the category →
Paintings by AI models: programs run through a simulation of oil paint.
AI · 21d ago · stillwet.art
PixVerse R2▲381A real-time world model you can explore and change
AI · 30d ago · world.pixverse.video
Source: Show HN post ↗ · launch text quoted in part
Launched alongside, January 2026
the whole month →- IN
Hey HN! I wanted to share something I built over the last few weeks: isometric.nyc is a massive isometric pixel art map of NYC, built with nano banana and coding agents. I didn't write a single line of code. Of course no-code doesn't mean no-engineering. This project took a lot more manual labor than I'd hoped! I wrote a deep dive on the workflow and some thoughts about the future of AI coding and creativity: http://cannoneyed.com/projects/isometric-nyc
AI · Jan 2026 · cannoneyed.com
Automatic AI-powered code reviews the moment you open a PR
Dev tools · Jan 2026 · kilo.ai
