[2025-04-28] 📚 Updates in #LRM

(1) Can Reasoning LLMs Enhance Clinical Document Classification?
(2) MultiMind: Enhancing Werewolf Agents with Multimodal Reasoning and Theory of Mind
(3) PropRAG: Guiding Retrieval with Beam Search over Proposition Paths

🔍 More at researchtrend.ai/communities/LRM

PropRAG swaps triples for sentence propositions and adds an LLM‑free beam search, reaching state‑of‑the‑art zero‑shot Recall@5 and F1 on 2Wiki, HotpotQA, MuSiQue, with code on GitHub. Read more: https://getnews.me/proprag-boosts-retrieval-with-beam-search-over-proposition-paths/ #proprag #retrieval

Jingjin Wang
PropRAG: Guiding Retrieval with Beam Search over Proposition Paths
https://arxiv.org/abs/2504.18070

PropRAG: Guiding Retrieval with Beam Search over Proposition Paths

Leverages contextually rich propositions and a novel beam search algorithm over proposition paths to explicitly discover multi-step reasoning chains.

📝 arxiv.org/abs/2504.18070
👨🏽‍💻 github.com/ReLink-Inc/P...

@rohanpaul_ai https://x.com/rohanpaul_ai/status/1921444660192661594 #x-rohanpaul_ai

PropRAG tries to solve for complex reasoning while retrieving passages using context-rich propositions.

It employs beam search over a proposition graph to find reasoning paths, avoiding online LLM co...

results on PopQA (55.3%), 2Wiki (93.7%), HotpotQA (97.0%), and MuSiQue (77.3%), alongside top F1 scores (e.g., 52.4% on MuSiQue). By improving evidence retrieval through richer representation and explicit, LLM-free online path finding, PropRAG [6/7 of https://arxiv.org/abs/2504.18070v1]

LLM inference costs and potential inconsistencies during evidence gathering. LLMs are used effectively offline for high-quality proposition extraction and post-retrieval for answer generation. PropRAG achieves state-of-the-art zero-shot Recall@5 [5/7 of https://arxiv.org/abs/2504.18070v1]

RAG methods like HippoRAG utilize knowledge graphs (KGs) built from triples, the inherent context loss limits fidelity. We introduce PropRAG, a framework leveraging contextually rich propositions and a novel beam search algorithm over proposition [3/7 of https://arxiv.org/abs/2504.18070v1]

Jingjin Wang: PropRAG: Guiding Retrieval with Beam Search over Proposition Paths https://arxiv.org/abs/2504.18070 https://arxiv.org/pdf/2504.18070 https://arxiv.org/html/2504.18070