Add side-by-side parser comparison on Anthropic Mythos PDF#4
Merged
Conversation
Responds to Karpathy's ask: "Post the converted Mythos pdf, figures, tables and all." Adds docs/examples/comparison/ with the same 245-page PDF run through seven PDF-to-markdown converters, plus a qualitative spot-check of one table (Table 6.3.A benchmark results) and one figure (Figure 2.2.5.2.A Virology Uplift Trial). Parsers compared: - nutrient CLI (this repo) - nutrient-sdk (Python, v1.0.4, premium) - docling 2.86.0 - pymupdf4llm 1.27.2.2 - markitdown[pdf] 0.1.5 - markit-ai 0.2.0 - @llamaindex/liteparse - pypdf 6.9.2 Key findings: 1. The free CLI in this repo matches docling and pymupdf4llm on structural extraction and is 15-75x faster. 2. Half the parsers (markitdown, pypdf, markit-ai, liteparse) produced zero markdown structure - all 245 pages collapsed into flat prose. 3. The Python SDK is the only parser that extracted figures. 110/110 figures in the Mythos PDF recovered as actual image files, in place, adjacent to their captions. The comparison directory includes: - README.md with the qualitative spot-check - One markdown file per parser (all 245 pages each) - figures/ with 110 images extracted by the Python SDK
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Responds to Karpathy's ask: "Post the converted Mythos pdf, figures, tables and all."
Adds
docs/examples/comparison/with the same 245-page Claude Mythos System Card run through seven PDF-to-markdown converters, plus a qualitative spot-check of one table and one figure.What's in the PR
docs/examples/comparison/README.mddocs/examples/comparison/anthropic-claude-mythos-*.mddocs/examples/comparison/figures/The canonical free-CLI output already lives one level up at
docs/examples/anthropic-claude-mythos-system-card.md.Key findings
Speed on this 245-page PDF
Structural extraction
Test plan
comparison/README.md