transcoder-circuits

A research tool for studying the internal circuits of large language models, which are neural networks that generate text. It uses transcoders to break parts of a model into smaller, more interpretable features and supports visual dashboards.

In plain words
What is it for?
Use it to train transcoders, analyze model sublayers, study interpretable features, compare them with sparse autoencoders, and test circuits through activation experiments.
Why use it?
It helps researchers investigate how a language model represents information and produces results instead of examining only its inputs and outputs.

Skill for Claude CodeCodex

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/zjunlp/mechanist/transcoder
Any agent
npx skills add zjunlp/Mechanist --skill transcoder
Clone the repo
git clone --depth 1 https://github.com/zjunlp/Mechanist

Made for: Claude Code, Codex.

Per session 50 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 5,378 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00050 $0.05378
Opus 5 $0.00025 $0.02689
Sonnet 5 $0.00010 $0.01076
Haiku 4.5 $0.00005 $0.00538

Measured 2d ago against content hash 8fa0c4780966, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

transcoder-circuits scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

The scan reads SKILL.md. This mod also ships 2 executable files (scripts/train_transcoder_example.py, scripts/transcoder_usage_example.py), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

skills/mechanism-skills/feature-dictionary-learning/transcoder/SKILL.md · 673 lines

How it starts

The opening of the file, as written. The whole thing — 673 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Transcoder Circuits: Reverse-Engineering LLM Circuits with Transcoders

When to Use

Activate this skill when:

  • Reverse-engineering circuits inside transformer language models (GPT-2, Pythia, etc.)
  • Training transcoders to decompose MLP sublayers into sparse linear combinations of features
  • Analyzing interpretable features in LLMs using sparse autoencoders or transcoders
  • Building feature dashboards or activation visualizations
  • Performing mechanistic interpretability research on neural networks
  • Comparing SAE vs. transcoder feature interpretability
  • Running circuit analysis, replacement contexts, or activation patching

Keywords: transcoder, SAE, sparse autoencoder, mechanistic interpretability, LLM circuits, MLP features, GPT-2, Pythia, feature dashboard, circuit analysis, activation patching

Quick Reference

Installation / Setup

Prerequisites

  • Python 3.8+
  • CUDA-capable GPU recommended

Quick Setup (from README)

bash setup.sh

This script installs dependencies and downloads transcoder weights from HuggingFace (pchlenski/gpt2-transcoders).

Manual Installation

pip install -r requirements.txt

Requirements (from requirements.txt)

Key dependencies include:

  • transformer_lens — for loading and hooking into transformer models
  • torch — PyTorch
  • einops
  • datasets
  • huggingface_hub
  • wandb (optional, for training logging)

Core Features

  • Transcoder Training: Train transcoders on LLM MLP sublayers to decompose activations into sparse interpretable features (sae_training/)
  • Circuit Analysis: Reverse-engineer fine-grained feature circuits within a model (transcoder_circuits/circuit_analysis.py)
  • Feature Dashboards: Generate dashboards for exploring transcoder and SAE features (transcoder_circuits/feature_dashboards.py)
  • Replacement Context: Swap MLP sublayers with transcoder reconstructions during inference (transcoder_circuits/replacement_ctx.py)
  • Activations Store: Stream tokens and generate/store activations during training (sae_training/activations_store.py)
  • Geometric Median: Utility for computing geometric median for initialization (sae_training/geom_median/)
  • SAE/Transcoder Comparison: Evaluation notebooks comparing SAEs and transcoders on Pythia-410M

Read the full file on GitHub · 673 lines

Files

What ships with it

4 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 673 lines · 50 tokens per session scan A 8fa0c4780966

Subscribe to this mod's changes

transcoder-circuits is a skill published in the GitHub repository zjunlp/Mechanist (51 stars, last pushed 6d ago), licensed MIT. It adds 50 tokens to every session and 5,378 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.