Gabriel Franco
66 posts
@gvsfranco
CS PhD student @BUCompSci. Interested in interpretability.
We don’t always know what problems are hard for LLMs. So devs evaluate on tasks HUMANS find hard or on broad benchmarks. What if we could instead anticipate which scenarios a model will fail on—all without evaluating specific input examples? 🧵NEW PAPER by @jenniferlumeng &al
🚨 New paper: Data-driven Circuit Discovery for Interpretability of Language Models 🚨 Do circuits actually explain how language models (LM) implement a task? In mechanistic interpretability, the goal of circuit study is to discover a “circuit” that is responsible for implementing a “task”. But we find that existing methods often discover circuits that are: ❌ not general task circuits: they do not capture the full range of mechanisms LMs uses across the task. Instead, they find: ✅ dataset-specific circuits: they explain how the model processes the examples used for circuit discovery. ✅ mixed-mechanism circuits: consisting of multiple independent mechanisms mixed in a single circuit. 1/🧵
🚨 New paper: Data-driven Circuit Discovery for Interpretability of Language Models 🚨 Do circuits actually explain how language models (LM) implement a task? In mechanistic interpretability, the goal of circuit study is to discover a “circuit” that is responsible for implementing a “task”. But we find that existing methods often discover circuits that are: ❌ not general task circuits: they do not capture the full range of mechanisms LMs uses across the task. Instead, they find: ✅ dataset-specific circuits: they explain how the model processes the examples used for circuit discovery. ✅ mixed-mechanism circuits: consisting of multiple independent mechanisms mixed in a single circuit. 1/🧵
Our paper "Latent Agents" was accepted to #ACL2026 Main! We distill multi-agent debate into a single LLM, matching debate performance at a fraction of the cost. We also show that internalized agents are discoverable and controllable. Huge thanks to @amuuueller and Dokyun Lee!
