Menu

RAG's Trust Problem: Citations, Memory, and Benchmarks Don't Add Up Quite Right
πŸ“°

RAG's Trust Problem: Citations, Memory, and Benchmarks Don't Add Up Quite Right

#devcommunity#dev#memory#agent#retrieval
Reading 0:00
15s threshold

This digest covers RAG/GraphRAG developments from 2026-11-09 through 2026-11-16: a dense week on arXiv and Hugging Face Daily Papers, a seven-post knowledge-layer push from Neo4j, and one methodology post each from LangChain and LlamaIndex.

πŸ”₯ Highlights

A 45-Minute Answer to One Question β€” graph-grounded context turns hours of log-diving into one query.

Cited but Unused: Parametric Provenance in Retrieval-Augmented Generation β€” up to 15.7% of RAG citations don't actually trace to retrieval.

MIRIX-2.0: Multi-Agent Memory System β€” new state of the art on LoCoMo and LongMemEval agent-memory benchmarks.

Introducing ParseBench-2 β€” a severity-weighted benchmark exposes where document parsers break RAG upstream.

DISCERN: Preserve at Write, Resolve at Read for Persistent LLM-Agent Memory β€” argues agent memory's real bottleneck is writing, not retrieving.

arXiv

RAG-FGO: Dynamic Optimization of Query-Rewriting Agents for Generative RAG β€” 2026-11-13 β€” Jointly optimizes discrete prompt templates and continuous…

Continue reading β€” create a free account

Join HashtagPLUS to read full articles, follow hashtags, vote, and join the conversation.