Newest papers

RESIN: Multi-Chain Reasoning with Entity-Structured Indexing

Xuanyu Meng, Jash Rajesh Parekh, Jiawei Han · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26782 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Argumentasi multi-chain adalah tugas yang menantang tidak hanya karena bukti multi-chain tersebar, tetapi juga karena bagian-bagian yang berbeda dari argumen mungkin memerlukan rantai dukungan yang berbeda, berinteraksi. Riset flat dapat membanjiri model dengan teks yang rusak, sementara pemilihan awal dari satu rute argumen dapat membuang bukti yang menjadi berguna hanya setelah fakta lain dipertimbangkan. Kami memperkenalkan RESIN (Reasoning with Entity Structured Indexing), yang menggunakan struktur untuk menemukan bukti sambil mempertahankan sumber asli sebagai dasar untuk argumen. RESIN memecah pertanyaan menjadi kebutuhan fokus satu-hop, mengembalikan satu set kecil unit berbasis sumber untuk setiap hop, dan hanya menempatkan bukti yang relevan ini dalam buffer alasan kompak. Buffer ini mempertahankan dukungan alternatif sampai komposisi akhir, membuat keputusan menengah secara eksplisit tanpa membuat jalur yang rapuh.

Machine translation (Meta M2M100); the English original follows.

Abstract

Multi-chain reasoning is a challenging task not only because the multi-chain evidence is dispersed, but also because different parts of reasoning may require distinct, interacting chains of support. Flat retrieval can flood a model with loosely related text, whereas early selection of a single reasoning path can discard evidence that becomes useful only after other facts are considered. We introduce RESIN (Reasoning with Entity Structured Indexing), which uses structure to locate evidence while keeping the original source as the basis for reasoning. RESIN decomposes a question into focused one-hop needs, retrieves a small set of source-grounded units for each hop, and places only this relevant evidence in a compact reasoning buffer. The buffer preserves alternative supports until final composition, making intermediate decisions explicit without forcing a brittle path. On the full HolisQA benchmark, RESIN achieves 82.92 context recall and 76.70 answer relevancy over all questions, with higher reported CR and AR across five domains and higher F1 in four domains than the published baselines. Ablations show that both typed evidence diversity and hop-wise decomposition are essential, supporting a clear conclusion: Precise evidence organization and delayed commitment make multi-chain reasoning effective and auditable. Our code is available at https://github.com/RamonMeng/RESIN.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Context (archaeology), Search engine indexing, Task (project management), Set (abstract data type), Recall

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange