GoCoMA: Hyperbolic Multimodal Representation Fusion for Large Language Model-Generated Code Attribution

2026-03-24

AI Summary

This work addresses the challenge of attributing code generated by large language models (LLMs) by proposing a novel multimodal provenance method that integrates source code style with visual representations of binary executable artifacts. It introduces hyperbolic space modeling to this task for the first time, employing Poincaré sphere embeddings to capture the hierarchical relationships between code and binary representations. A geodesic cosine similarity-based cross-modal attention mechanism (GCSA) is designed to effectively fuse these heterogeneous modalities, followed by a back-projection into Euclidean space to enhance discriminability. Evaluated on the CoDET-M4 and LLMAuthorBench benchmarks, the proposed approach significantly outperforms existing unimodal and Euclidean multimodal baselines, demonstrating its effectiveness and innovation in LLM-generated code attribution.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Computer Vision: Multi-modal Vision

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Large Language Models (LLMs) trained on massive code corpora are now increasingly capable of generating code that is hard to distinguish from human-written code. This raises practical concerns, including security vulnerabilities and licensing ambiguity, and also motivates a forensic question:'Who (or which LLM) wrote this piece of code?'We present GoCoMA, a multimodal framework that models an extrinsic hierarchy between (i) code stylometry, capturing higher-level structural and stylistic signatures, and (ii) image representations of binary pre-executable artifacts (BPEA), capturing lower-level, execution-oriented byte semantics shaped by compilation and toolchains. GoCoMA projects modality embeddings into a hyperbolic Poincar\'e ball, fuses them via a geodesic-cosine similarity-based cross-modal attention (GCSA) fusion mechanism, and back-projects the fused representation to Euclidean space for final LLM-source attribution. Experiments on two open-source benchmarks (CoDET-M4 and LLMAuthorBench) show that GoCoMA consistently outperforms unimodal and Euclidean multimodal baselines under identical evaluation protocols.

Problem

• code attribution
• large language models
• code stylometry
• binary artifacts
• multimodal representation

Innovation

• hyperbolic representation
• multimodal fusion
• code attribution
• geodesic-cosine similarity
• stylometry