This page collects practical resources and guidelines for students and collaborators working with me — e.g. how to upload papers to arXiv, information about the conferences we typically submit to, and other lab conventions. It is a work in progress and will be filled in over time.

We typically write papers in Overleaf. arXiv expects submission files in a specific format, and it's easy to get stuck on non-compiling submissions. Suggested workflow:

1. Use the arXiv LaTeX cleaner (pip install arxiv-latex-cleaner, or on Mac brew install arxiv_latex_cleaner) before you publish — otherwise your comments, unused files, etc. will be visible to the public.
2. On Overleaf, click "Submit" → arXiv → "Download project ZIP".
3. Unzip the downloaded file, then run arxiv_latex_cleaner PROJECT_DIRECTORY in the command line (you may need the --keep_bib option). This creates a cleaned directory with an _arXiv suffix. Note: this automatically resizes images to save space; if you have high-resolution figures that need to be displayed at full resolution, adjust this in the cleaner's options.
4. If your project has nested directories, you may need to move the bibliography file (*.bib) into the inner directory.
5. Zip the cleaned directory and submit it to arXiv (arXiv requires a zipped directory for upload).

We mostly target the major NLP, vision, and speech venues, including:

• NLP: ACL, EMNLP, NAACL, and ACL Rolling Review (ARR), plus relevant workshops (e.g. ArabicNLP)
• Vision: CVPR, ICCV, ECCV, NeurIPS, ICLR, WACV
• Speech: Interspeech, ICASSP, SLT
• Other: Eurographics, ICDAR

Useful trackers for deadlines:

• AI Deadlines — most relevant categories are ML, CV, and NLP.

Before submitting, it's worth getting feedback on a draft: Stanford Agentic Reviewer gives free AI-generated review feedback (you can specify the target venue) to help catch issues ahead of the real reviewers.

Once a paper is accepted (or posted on arXiv), it's standard to prepare a project page. Feel free to reuse the format of this site or of our papers' project pages (see the Publications page) — you can "view source" in your browser to get the HTML and adapt it, and host it for free with GitHub Pages.

Posters:

• Good tools: Microsoft PowerPoint or draw.io.
• Itai Lang's poster guide.
• CVPR & NeurIPS poster templates (Bolei Zhou).

Videos: some conferences require a short video presenting your work.

• You can record slides with narration via Zoom or PowerPoint's recording feature.
• Recommended: use Adobe Enhance (free) to clean up audio quality. Split the audio track with ffmpeg (ffmpeg -i video.mp4 video.mp3), enhance it, then splice it back in (ffmpeg -i video.mp4 -i enhanced_audio.wav -c:v copy -map 0:v:0 -map 1:a:0 new_video.mp4).

Finding and organizing papers:

• Google Scholar
• Connected Papers
• Zotero / Mendeley for reference management
• Scholar Inbox for personalized paper recommendations

Staying up to date:

• Deep Learning Monitor

Including images in papers/talks: make sure images have a free license — use Google Image Search's "Usage Rights" filter, or search Wikimedia Commons or Unsplash (and follow best practices for attribution).