Commit 49ae51a
fix(preprints): render LaTeX math in automated-review reports
Reviewers reading LaTeX papers emit inline/display math (e.g.
$\Delta_{\cos} \in [0.02, 0.05]$); tex_escape was turning it into literal
source. Now math spans ($...$, $$...$$, \(..\), \[..\]) pass through verbatim
and compile as math (amsmath/amssymb added to the review preamble).
Robustness:
- Single-$ spans are guarded: a span containing a space-delimited prose word is
kept literal, since reviewers sometimes quote a paper's UNBALANCED $ (which
would otherwise pull a whole clause into math mode).
- build_preprint_pdfs compiles with math, and retries once with math-as-text if
a stray/paper-defined macro breaks lualatex — one bad expression never costs
the whole report (fired once across the 60-preprint rebuild; all 60 built).
Rebuilt all 60 preprints. Tests added for math passthrough + fallback.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>1 parent eb05f2a commit 49ae51a
65 files changed
Lines changed: 91 additions & 21 deletions
File tree
- projects
- PROJ-562-a-stylometric-application-of-large-langu/paper/pdf
- PROJ-563-many-shot-cot-icl-making-in-context-lear/paper/pdf
- PROJ-564-qwen-image-vae-2-0-technical-report/paper/pdf
- PROJ-565-edit-compass-editreward-compass-a-unifie/paper/pdf
- PROJ-566-mint-managed-infrastructure-for-training/paper/pdf
- PROJ-567-anyflow-any-step-video-diffusion-model-w/paper/pdf
- PROJ-568-identifying-stimulus-driven-neural-activ/paper/pdf
- PROJ-569-intern-atlas-a-methodological-evolution/paper/pdf
- PROJ-570-leveraging-verifier-based-reinforcement/paper/pdf
- PROJ-571-co-evolving-policy-distillation/paper/pdf
- PROJ-572-https-arxiv-org-abs-2604-28185/paper/pdf
- PROJ-573-https-arxiv-org-abs-2604-27351/paper/pdf
- PROJ-574-eva-bench-a-new-end-to-end-framework-for/paper/pdf
- PROJ-575-training-long-context-vision-language-mo/paper/pdf
- PROJ-576-sana-wm-efficient-minute-scale-world-mod/paper/pdf
- PROJ-577-multabench-benchmarking-multimodal-tabul/paper/pdf
- PROJ-578-https-arxiv-org-abs-2605-14906/paper/pdf
- PROJ-579-https-arxiv-org-abs-2605-15155/paper/pdf
- PROJ-580-https-arxiv-org-abs-2605-15141/paper/pdf
- PROJ-597-https-arxiv-org-abs-2605-11739/paper/pdf
- PROJ-598-https-arxiv-org-abs-2605-15824/paper/pdf
- PROJ-601-https-arxiv-org-abs-2605-12882/paper/pdf
- PROJ-602-https-arxiv-org-abs-2605-18661/paper/pdf
- PROJ-604-https-arxiv-org-abs-2605-18739/paper/pdf
- PROJ-607-https-arxiv-org-abs-2605-19769/paper/pdf
- PROJ-608-autoresearchclaw-self-reinforcing-autono/paper/pdf
- PROJ-612-a-survey-of-large-audio-language-models/paper/pdf
- PROJ-615-mega-asr-towards-in-the-wild-2-speech-re/paper/pdf
- PROJ-616-video2gui-synthesizing-large-scale-inter/paper/pdf
- PROJ-622-https-arxiv-org-abs-2605-23463/paper/pdf
- PROJ-623-sciatlas-a-large-scale-knowledge-graph-f/paper/pdf
- PROJ-628-foundation-protocol-a-coordination-layer/paper/pdf
- PROJ-631-https-arxiv-org-abs-2605-25604/paper/pdf
- PROJ-633-mobilegym-a-verifiable-and-highly-parall/paper/pdf
- PROJ-636-locateanything-fast-and-high-quality-vis/paper/pdf
- PROJ-639-https-arxiv-org-abs-2605-28774/paper/pdf
- PROJ-648-representation-forcing-for-bottleneck-fr/paper/pdf
- PROJ-656-crafter-a-multi-agent-harness-for-editab/paper/pdf
- PROJ-662-https-arxiv-org-abs-2606-03746/paper/pdf
- PROJ-664-https-arxiv-org-abs-2606-02060/paper/pdf
- PROJ-684-latent-spatial-memory-for-video-world-mo/paper/pdf
- PROJ-688-agents-last-exam/paper/pdf
- PROJ-703-evoarena-tracking-memory-evolution-for-r/paper/pdf
- PROJ-707-appo-agentic-procedural-policy-optimizat/paper/pdf
- PROJ-708-omnidirector-general-multi-shot-camera-c/paper/pdf
- PROJ-720-joyai-vl-interaction-real-time-vision-la/paper/pdf
- PROJ-730-ace-ego-0-unifying-egocentric-human-and/paper/pdf
- PROJ-732-learning-from-the-self-future-on-policy/paper/pdf
- PROJ-739-guava-an-effective-and-universal-harness/paper/pdf
- PROJ-741-molmomotion-forecasting-point-trajectori/paper/pdf
- PROJ-748-multi-lcb-extending-livecodebench-to-mul/paper/pdf
- PROJ-750-dragmesh-2-physically-plausible-dexterou/paper/pdf
- PROJ-751-moebius-0-2b-lightweight-image-inpaintin/paper/pdf
- PROJ-767-gatemem-benchmarking-memory-governance-i/paper/pdf
- PROJ-768-memslides-a-hierarchical-memory-driven-a/paper/pdf
- PROJ-769-perceptiondlm-parallel-region-perception/paper/pdf
- PROJ-788-improved-large-language-diffusion-models/paper/pdf
- PROJ-789-shuttermuse-capture-time-photography-gui/paper/pdf
- PROJ-801-translation-as-a-bridging-action-transfe/paper/pdf
- PROJ-804-formalizing-latent-thoughts-four-axioms/paper/pdf
- src/llmxive/paper_reprocess
- tests/unit
Some content is hidden
Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
Binary file not shown.
0 commit comments