frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: Symbolic Circuit Distillation: prove program to LLM circuit equivalence

https://github.com/neelsomani/symbolic-circuit-distillation
16•nsomani•1mo ago
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 behavior (e.g. quote closing or bracket depth) extracted from a transformer. 2. Treat the circuit as an executable function and train a tiny ReLU network ("surrogate") that exactly matches the circuit on all inputs in a bounded domain (typically sequences of length 5–10 over a small token alphabet). 3. Search over a constrained DSL of common transformer motifs (counters, toggles, threshold detectors, small state machines) to synthesize candidate Python programs. 4. Use SMT-based bounded equivalence checking to either: - Prove that a candidate program and the surrogate agree on all inputs in the domain, or - Produce a counterexample input that rules the program out.

If the solver finds a proof, you get a small, human-readable Python function plus a machine-checkable guarantee that it matches the original circuit on that bounded domain.

Why I built this

Mechanistic interpretability has gotten pretty good at extracting "small crisp circuits" from large models, but turning those graphs into clean, human-readable algorithms is still very manual. My goal here is to automate that last step: go from "here is a sparse circuit" to "here is a verified algorithm that explains what it does", without hand-holding.

What works today

- Tasks: quote closing and bracket-depth detection from the OpenAI circuit_sparsity repo. - Exact surrogate fitting on a finite token domain. - DSL templates for simple counters, toggles, and small state machines. - SMT-based bounded equivalence between: sparse circuit -> ReLU surrogate -> Python program in the DSL.

Limitations and open questions

- The guarantees are bounded: equivalence is only proven on a finite token domain (short sequences and a small vocabulary). - Currently focused on very small circuits. Scaling to larger circuits and longer contexts is open engineering and research work. - The DSL is hand-designed around a few motifs. I am not yet learning the DSL itself or doing anything very clever in the search.

What I would love feedback on

- Are the problem framing and guarantees interesting to people working on mechanistic interpretability or formal methods? - Suggestions for next benchmarks: which circuits or behaviors would you want to see distilled next? - Feedback on the DSL design, search strategy, and SMT setup.

Happy to answer questions about implementation details, the SMT encoding, integration with OpenAI's Sparse Circuits repo, or anything else.

Comments

aappleby•1mo ago
No examples in the readme?
nsomani•1mo ago
There are two examples provided - quote matching and bracket closing.

Female Asian Elephant Calf Born at the Smithsonian National Zoo

https://www.si.edu/newsdesk/releases/female-asian-elephant-calf-born-smithsonians-national-zoo-an...
1•gmays•1m ago•0 comments

Show HN: Zest – A hands-on simulator for Staff+ system design scenarios

https://staff-engineering-simulator-880284904082.us-west1.run.app/
1•chanip0114•2m ago•1 comments

Show HN: DeSync – Decentralized Economic Realm with Blockchain-Based Governance

https://github.com/MelzLabs/DeSync
1•0xUnavailable•6m ago•0 comments

Automatic Programming Returns

https://cyber-omelette.com/posts/the-abstraction-rises.html
1•benrules2•9m ago•1 comments

Why Are There Still So Many Jobs? The History and Future of Workplace Automation [pdf]

https://economics.mit.edu/sites/default/files/inline-files/Why%20Are%20there%20Still%20So%20Many%...
2•oidar•12m ago•0 comments

The Search Engine Map

https://www.searchenginemap.com
1•cratermoon•19m ago•0 comments

Show HN: Souls.directory – SOUL.md templates for AI agent personalities

https://souls.directory
1•thedaviddias•20m ago•0 comments

Real-Time ETL for Enterprise-Grade Data Integration

https://tabsdata.com
1•teleforce•24m ago•0 comments

Economics Puzzle Leads to a New Understanding of a Fundamental Law of Physics

https://www.caltech.edu/about/news/economics-puzzle-leads-to-a-new-understanding-of-a-fundamental...
2•geox•25m ago•0 comments

Switzerland's Extraordinary Medieval Library

https://www.bbc.com/travel/article/20260202-inside-switzerlands-extraordinary-medieval-library
2•bookmtn•25m ago•0 comments

A new comet was just discovered. Will it be visible in broad daylight?

https://phys.org/news/2026-02-comet-visible-broad-daylight.html
2•bookmtn•30m ago•0 comments

ESR: Comes the news that Anthropic has vibecoded a C compiler

https://twitter.com/esrtweet/status/2019562859978539342
1•tjr•31m ago•0 comments

Frisco residents divided over H-1B visas, 'Indian takeover' at council meeting

https://www.dallasnews.com/news/politics/2026/02/04/frisco-residents-divided-over-h-1b-visas-indi...
1•alephnerd•32m ago•0 comments

If CNN Covered Star Wars

https://www.youtube.com/watch?v=vArJg_SU4Lc
1•keepamovin•38m ago•0 comments

Show HN: I built the first tool to configure VPSs without commands

https://the-ultimate-tool-for-configuring-vps.wiar8.com/
2•Wiar8•41m ago•3 comments

AI agents from 4 labs predicting the Super Bowl via prediction market

https://agoramarket.ai/
1•kevinswint•46m ago•1 comments

EU bans infinite scroll and autoplay in TikTok case

https://twitter.com/HennaVirkkunen/status/2019730270279356658
5•miohtama•48m ago•3 comments

Benchmarking how well LLMs can play FizzBuzz

https://huggingface.co/spaces/venkatasg/fizzbuzz-bench
1•_venkatasg•51m ago•1 comments

Why I Joined OpenAI

https://www.brendangregg.com/blog/2026-02-07/why-i-joined-openai.html
19•SerCe•51m ago•11 comments

Octave GTM MCP Server

https://docs.octavehq.com/mcp/overview
1•connor11528•53m ago•0 comments

Show HN: Portview what's on your ports (diagnostic-first, single binary, Linux)

https://github.com/Mapika/portview
3•Mapika•55m ago•0 comments

Voyager CEO says space data center cooling problem still needs to be solved

https://www.cnbc.com/2026/02/05/amazon-amzn-q4-earnings-report-2025.html
1•belter•58m ago•0 comments

Boilerplate Tax – Ranking popular programming languages by density

https://boyter.org/posts/boilerplate-tax-ranking-popular-languages-by-density/
1•nnx•59m ago•0 comments

Zen: A Browser You Can Love

https://joeblu.com/blog/2026_02_zen-a-browser-you-can-love/
1•joeblubaugh•1h ago•0 comments

My GPT-5.3-Codex Review: Full Autonomy Has Arrived

https://shumer.dev/gpt53-codex-review
2•gfortaine•1h ago•0 comments

Show HN: FastLog: 1.4 GB/s text file analyzer with AVX2 SIMD

https://github.com/AGDNoob/FastLog
2•AGDNoob•1h ago•1 comments

God said it (song lyrics) [pdf]

https://www.lpmbc.org/UserFiles/Ministries/AVoices/Docs/Lyrics/God_Said_It.pdf
1•marysminefnuf•1h ago•0 comments

I left Linus Tech Tips [video]

https://www.youtube.com/watch?v=gqVxgcKQO2E
1•ksec•1h ago•0 comments

Program Theory

https://zenodo.org/records/18512279
1•Anonymus12233•1h ago•0 comments

Show HN: Local DNA analysis skill for OpenClaw

https://github.com/wkyleg/personal-genomics
2•wkyleg•1h ago•0 comments