@inproceedings{brunello-etal-2025-blackboxnlp,
title = "{B}lackbox{NLP}-2025 {MIB} Shared Task: {IPE}: Isolating Path Effects for Improving Latent Circuit Identification",
author = "Brunello, Nicol{\`o} and
Cerutti, Andrea and
Sassella, Andrea and
Carman, Mark James",
editor = "Belinkov, Yonatan and
Mueller, Aaron and
Kim, Najoung and
Mohebbi, Hosein and
Chen, Hanjie and
Arad, Dana and
Sarti, Gabriele",
booktitle = "Proceedings of the 8th {B}lackbox{NLP} Workshop: Analyzing and Interpreting Neural Networks for {NLP}",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.blackboxnlp-1.30/",
doi = "10.18653/v1/2025.blackboxnlp-1.30",
pages = "528--536",
ISBN = "979-8-89176-346-3",
abstract = "Understanding why large language models (LLMs) exhibit certain behaviors is the goal of mechanistic interpretability. One of the major tools employed by mechanistic interpretability is circuit discovery, i.e., identifying a subset of the model{'}s components responsible for a given task. We present a novel circuit discovery technique called IPE (Isolating Path Effects) that, unlike traditional edge-centric approaches, aims to identify entire computational paths (from input embeddings to output logits) responsible for certain model behaviors. Our method modifies the messages passed between nodes along a given path in such a way as to either precisely remove the effects of the entire path (i.e., ablate it) or to replace the path{'}s effects with those that would have been generated by a counterfactual input. IPE is different from current path-patching or edge activation-patching techniques since they are not ablating single paths, but rather a set of paths sharing certain edges, preventing more precise tracing of information flow. We apply our method to the well-known Indirect Object Identification (IOI) task, recovering the canonical circuit reported in prior work. On the MIB workshop leaderboard, we tested IOI and MCQA tasks on GPT2-small and Qwen2.5. For GPT2, path counterfactual replacement outperformed path ablation as expected and led to top-ranking results, while for Qwen, no significant differences were observed, indicating a need for larger experiments to distinguish the two approaches."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="brunello-etal-2025-blackboxnlp">
<titleInfo>
<title>BlackboxNLP-2025 MIB Shared Task: IPE: Isolating Path Effects for Improving Latent Circuit Identification</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nicolò</namePart>
<namePart type="family">Brunello</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrea</namePart>
<namePart type="family">Cerutti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrea</namePart>
<namePart type="family">Sassella</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mark</namePart>
<namePart type="given">James</namePart>
<namePart type="family">Carman</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 8th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yonatan</namePart>
<namePart type="family">Belinkov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aaron</namePart>
<namePart type="family">Mueller</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Najoung</namePart>
<namePart type="family">Kim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hosein</namePart>
<namePart type="family">Mohebbi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hanjie</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dana</namePart>
<namePart type="family">Arad</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gabriele</namePart>
<namePart type="family">Sarti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-346-3</identifier>
</relatedItem>
<abstract>Understanding why large language models (LLMs) exhibit certain behaviors is the goal of mechanistic interpretability. One of the major tools employed by mechanistic interpretability is circuit discovery, i.e., identifying a subset of the model’s components responsible for a given task. We present a novel circuit discovery technique called IPE (Isolating Path Effects) that, unlike traditional edge-centric approaches, aims to identify entire computational paths (from input embeddings to output logits) responsible for certain model behaviors. Our method modifies the messages passed between nodes along a given path in such a way as to either precisely remove the effects of the entire path (i.e., ablate it) or to replace the path’s effects with those that would have been generated by a counterfactual input. IPE is different from current path-patching or edge activation-patching techniques since they are not ablating single paths, but rather a set of paths sharing certain edges, preventing more precise tracing of information flow. We apply our method to the well-known Indirect Object Identification (IOI) task, recovering the canonical circuit reported in prior work. On the MIB workshop leaderboard, we tested IOI and MCQA tasks on GPT2-small and Qwen2.5. For GPT2, path counterfactual replacement outperformed path ablation as expected and led to top-ranking results, while for Qwen, no significant differences were observed, indicating a need for larger experiments to distinguish the two approaches.</abstract>
<identifier type="citekey">brunello-etal-2025-blackboxnlp</identifier>
<identifier type="doi">10.18653/v1/2025.blackboxnlp-1.30</identifier>
<location>
<url>https://aclanthology.org/2025.blackboxnlp-1.30/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>528</start>
<end>536</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T BlackboxNLP-2025 MIB Shared Task: IPE: Isolating Path Effects for Improving Latent Circuit Identification
%A Brunello, Nicolò
%A Cerutti, Andrea
%A Sassella, Andrea
%A Carman, Mark James
%Y Belinkov, Yonatan
%Y Mueller, Aaron
%Y Kim, Najoung
%Y Mohebbi, Hosein
%Y Chen, Hanjie
%Y Arad, Dana
%Y Sarti, Gabriele
%S Proceedings of the 8th BlackboxNLP Workshop: Analyzing and Interpreting Neural Networks for NLP
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-346-3
%F brunello-etal-2025-blackboxnlp
%X Understanding why large language models (LLMs) exhibit certain behaviors is the goal of mechanistic interpretability. One of the major tools employed by mechanistic interpretability is circuit discovery, i.e., identifying a subset of the model’s components responsible for a given task. We present a novel circuit discovery technique called IPE (Isolating Path Effects) that, unlike traditional edge-centric approaches, aims to identify entire computational paths (from input embeddings to output logits) responsible for certain model behaviors. Our method modifies the messages passed between nodes along a given path in such a way as to either precisely remove the effects of the entire path (i.e., ablate it) or to replace the path’s effects with those that would have been generated by a counterfactual input. IPE is different from current path-patching or edge activation-patching techniques since they are not ablating single paths, but rather a set of paths sharing certain edges, preventing more precise tracing of information flow. We apply our method to the well-known Indirect Object Identification (IOI) task, recovering the canonical circuit reported in prior work. On the MIB workshop leaderboard, we tested IOI and MCQA tasks on GPT2-small and Qwen2.5. For GPT2, path counterfactual replacement outperformed path ablation as expected and led to top-ranking results, while for Qwen, no significant differences were observed, indicating a need for larger experiments to distinguish the two approaches.
%R 10.18653/v1/2025.blackboxnlp-1.30
%U https://aclanthology.org/2025.blackboxnlp-1.30/
%U https://doi.org/10.18653/v1/2025.blackboxnlp-1.30
%P 528-536
Markdown (Informal)
[BlackboxNLP-2025 MIB Shared Task: IPE: Isolating Path Effects for Improving Latent Circuit Identification](https://aclanthology.org/2025.blackboxnlp-1.30/) (Brunello et al., BlackboxNLP 2025)
ACL