---
type: "raw_signal"
id: "basket-2459512c-31"
slug: "agent-document-data-stack-roundup"
title: "Agent Document Stacks Split Parsing, Embeddings, and Evaluation"
description: "New releases show document agents becoming composed systems with separate extraction, representation, and quality layers."
retrieval_nugget: "New releases show document agents becoming composed systems with separate extraction, representation, and quality layers. A cluster of releases is separating the document-agent stack into parsing, multimodal representation, and evaluation."
observed_at: "2026-08-28"
record_date: "2026-08-28"
date_kind: "observed_at"
why_it_matters: "The practical architecture is routed: classify the document, choose an extraction path, validate a schema, embed what must remain searchable, and evaluate failures on representative files. The roundup remains an implementation radar, not proof that one stack is best; each component still needs task-specific testing and provenance."
novelty: "structural"
verification_level: "source-linked"
signal_type: "roundup"
source_platform: "x.com"
topics: ["document-ai","embeddings","evals"]
entities: []
related_patterns: []
source_url: "https://x.com/jerryjliu0/status/2093028218467172357"
source_urls: ["https://x.com/jerryjliu0/status/2093028218467172357","https://x.com/cohere/status/2092962407849050478","https://x.com/cohere/status/2092962411284189285","https://x.com/cohere/status/2092962413154832671","https://llamaindex.ai/blog/introducing-extractbench","https://github.com/Tencent/WeMM-Embedding"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "signal:agent-document-data-stack-roundup"
status: "published"
visuals: []
routes: {"html":"https://newruntime.com/signals/agent-document-data-stack-roundup/","markdown":"https://newruntime.com/signals/agent-document-data-stack-roundup.md","json":"https://newruntime.com/signals/agent-document-data-stack-roundup.json"}
---

# Agent Document Stacks Split Parsing, Embeddings, and Evaluation

## Retrieval answer

New releases show document agents becoming composed systems with separate extraction, representation, and quality layers. A cluster of releases is separating the document-agent stack into parsing, multimodal representation, and evaluation.

A cluster of releases is separating the document-agent stack into parsing, multimodal representation, and evaluation.

LlamaIndex's ExtractBench makes extraction quality a distinct test surface, while WeMM-Embedding and related releases focus on how mixed document content is represented for retrieval and downstream use. Read together, they make one-model document processing look increasingly incomplete.

The practical architecture is routed: classify the document, choose an extraction path, validate a schema, embed what must remain searchable, and evaluate failures on representative files. The roundup remains an implementation radar, not proof that one stack is best; each component still needs task-specific testing and provenance.