PulseAugur
LIVE 06:49:23
research · [4 sources] ·
0
research

New research tackles AI agent citation accuracy and document format efficiency

A new evaluation framework, "Cited but Not Verified," has been developed to assess the source attribution capabilities of large language models (LLMs) used in research agents. This framework parses and evaluates inline citations from LLM-generated reports across three dimensions: link accessibility, content relevance, and factual accuracy. Benchmarking 14 LLMs revealed that while frontier models maintain high link validity and relevance, their factual accuracy in citations is significantly lower, especially as the depth of retrieval increases. Separately, a new file format called ObjectGraph (.og) has been proposed to address the inefficiencies of current document handling by LLM agents, reconceiving documents as traversable knowledge graphs rather than linear text. AI

Summary written by gemini-2.5-flash-lite from 4 sources. How we write summaries →

IMPACT New evaluation frameworks and file formats are emerging to improve the reliability and efficiency of LLM agents in research and information synthesis.

RANK_REASON The cluster contains two academic papers detailing new frameworks and file formats for LLM research.

Read on arXiv cs.CL →

COVERAGE [4]

  1. arXiv cs.CL TIER_1 · Hailey Onweller, Elias Lumer, Austin Huber, Pia Ramchandani, Vamse Kumar Subbiah, Corey Feld ·

    Cited but Not Verified: Parsing and Evaluating Source Attribution in LLM Deep Research Agents

    arXiv:2605.06635v1 Announce Type: new Abstract: Large language models (LLMs) power deep research agents that synthesize information from hundreds of web sources into cited reports, yet these citations cannot be reliably verified. Current approaches either trust models to self-cit…

  2. arXiv cs.CL TIER_1 · Corey Feld ·

    Cited but Not Verified: Parsing and Evaluating Source Attribution in LLM Deep Research Agents

    Large language models (LLMs) power deep research agents that synthesize information from hundreds of web sources into cited reports, yet these citations cannot be reliably verified. Current approaches either trust models to self-cite accurately, risking bias, or employ retrieval-…

  3. arXiv cs.AI TIER_1 · Mohit Dubey, Open Gigantic ·

    ObjectGraph: From Document Injection to Knowledge Traversal -- A Native File Format for the Agentic Era

    arXiv:2604.27820v1 Announce Type: new Abstract: Every document format in existence was designed for a human reader moving linearly through text. Autonomous LLM agents do not read - they retrieve. This fundamental mismatch forces agents to inject entire documents into their contex…

  4. arXiv cs.AI TIER_1 · Open Gigantic ·

    ObjectGraph: From Document Injection to Knowledge Traversal -- A Native File Format for the Agentic Era

    Every document format in existence was designed for a human reader moving linearly through text. Autonomous LLM agents do not read - they retrieve. This fundamental mismatch forces agents to inject entire documents into their context window, wasting tokens on irrelevant content, …