paper-with-me

홈 › Papers

CogDoc: Towards Unified thinking in Documents

2025-12-14 · Qixin Xu, Haozhe Wang, Che Liu, Fangzhen Lin, Wenhu Chen arxiv

Current document reasoning paradigms are constrained by a fundamental trade-off between scalability (processing long-context documents) and fidelity (capturing fine-grained, multimodal details). To bridge this gap, we propose CogDoc, a unified coarse-to-fine thinking framework that mimics human cognitive processes: a low-resolution "Fast Reading" phase for scalable information localization,followed by a high-resolution "Focused Thinking" phase for deep reasoning. We conduct a rigorous investigation into post-training strategies for the unified thinking framework, demonstrating that a Direct Reinforcement Learning (RL) approach outperforms RL with Supervised Fine-Tuning (SFT) initialization. Specifically, we find that direct RL avoids the "policy conflict" observed in SFT. Empirically, our 7B model achieves state-of-the-art performance within its parameter class, notably surpassing significantly larger proprietary models (e.g., GPT-4o) on challenging, visually rich document benchmarks.

📄 PDF Abstract BibTeX arXiv:2512.12658

Code (0)

등록된 구현이 없습니다.

Tasks

Reinforcement Learning

Similar Papers 제목 키워드 기반

Learning More Effective Representations for Dense Retrieval through Deliberate Thinking Before Search

2025-02-18 · Yifan Ji, Zhipeng Xu, Zhenghao Liu, Yukun Yan 외

Recent dense retrievers usually thrive on the emergency capabilities of Large Language Models (LLMs), using them to encode queries and documents into an embedding space for retrieval. These LLM-based dense retrievers hav…

Retrieval

IRGAN: A Minimax Game for Unifying Generative and Discriminative Information Retrieval Models

2017-05-30 · Jun Wang, Lantao Yu, Wei-Nan Zhang, Yu Gong 외

This paper provides a unified account of two schools of thinking in information retrieval modelling: the generative retrieval focusing on predicting relevant documents given a query, and the discriminative retrieval focu…

Ad-Hoc Information RetrievalDocument RankingInformation RetrievalQuestion Answering+1

OptimalThinkingBench: Evaluating Over and Underthinking in LLMs

2025-08-18 · Pranjal Aggarwal, Seungone Kim, Jack Lanchantin, Sean Welleck 외 arxiv

Thinking LLMs solve complex tasks at the expense of increased compute and overthinking on simpler problems, while non-thinking LLMs are faster and cheaper but underthink on harder reasoning problems. This has led to the …

MM-BizRAG: Rethinking Multimodal Retrieval-Augmented Generation for General Purpose Enterprise Q&A

2026-06-02 · Hanoz Bhathena, Parin Rajesh Jhaveri, Rohan Mittal, Prateek Singh 외 arxiv

Recent advances in multimodal retrieval-augmented generation (MM-RAG) have shifted toward minimal parsing, relying on page-level images for producing retriever embeddings and for answer generation. While efficient, this …

Answer Generation

Thinking with Video: Video Generation as a Promising Multimodal Reasoning Paradigm

2025-11-06 · Jingqi Tong, Yurong Mou, Hangcheng Li, Mingzhe Li 외 arxiv

The "Thinking with Text" and "Thinking with Images" paradigms significantly improve the reasoning abilities of large language models (LLMs) and Vision-Language Models (VLMs). However, these paradigms have inherent limita…

Multimodal ReasoningVideo Generation