Research repositories from Layer 6 AI
Open research code with real-world impact.
This group-level landing page summarizes Layer 6's public open source activity across completed research implementations, active projects, project pages, benchmarks, and conference-linked code releases.
Snapshot
A research portfolio built for reuse.
Data snapshot: public GitHub API for layer6ai-labs, refreshed July 15, 2026.
Open source activity summary
A company-level overview, not an individual project README.
Layer 6 uses its public GitHub organization to share reusable research artifacts with the broader machine learning community. This page summarizes that activity across the organization so evaluators can inspect the open source portfolio in one place.
Published research implementations
Mature releases include DropoutNet, RecSys2018, T-Fixup, GSS, HGCF, ProxyFL, dgm-eval, and xpool. These projects document completed research contributions with code, evaluation routines, or project pages.
Recent and active public work
Current public activity includes forest-diffusion-mo, CCPFN-inference, TabDPT-inference, Agentic-Monte-Carlo, RankJudge, DRESS, and conformal-agent-error-attribution, with several pushes in June and July 2026.
Reusable artifacts for applied ML
Open source helps Layer 6 turn research into inspectable code, reproducible baselines, model inference packages, benchmark tooling, and evaluation assets that can be reviewed and extended by researchers, engineers, and collaborators.
Conference-linked releases and project sites
Public repository descriptions and linked pages connect Layer 6 code to venues and communities such as NeurIPS, ICML, ICLR, CVPR, ICCV, WWW, and RecSys, alongside GitHub Pages sites for selected projects.
Contribution areas
Layer 6 releases work where reproducibility matters.
The organization spans mature academic codebases, active benchmark work, model inference packages, and project documentation that helps the broader community inspect and extend Layer 6 research.
Tabular and causal foundation models
Inference and training resources for TabDPT, retrieval and fine-tuning work for PFN-style tabular models, and causal foundation model inference.
- TabDPT-inference
- TabDPT-training
- LoCalPFN
- CCPFN-inference
Generative modeling and evaluation
Tooling for evaluating deep generative models, geometry-focused project pages, and scalable diffusion or flow-matching implementations.
- dgm-eval
- dgm_geometry
- calo-forest
- rectangular-flows
Recommenders and ranking
Long-running recommender systems contributions, challenge-winning code, collaborative filtering models, and ranking evaluation research.
- DropoutNet
- RecSys2018
- HGCF
- RankJudge
Agents, RAG, and language systems
Recent work on agent simulation, RAG error classification, conformal factuality, text-to-SQL, policy optimization, and conference generation.
- Agentic-Monte-Carlo
- conformal-rag
- rag-error-classification
- wapo
Vision and multimodal learning
Open implementations for image retrieval, scene graph generation, video-language retrieval, multimodal fusion, and action selection.
- xpool
- GSS
- fusemix
- SGG-Seq2Seq
Reliability, fairness, and federated learning
Shared code around conformal prediction, fairness evaluation, proxy model sharing, and attribution of failures in agentic systems.
- ProxyFL
- llm-in-the-loop-conformal-fairness
- conformal-agent-error-attribution
- textual-bayes
Featured repositories
Completed and widely reused public projects.
These completed or mature repositories lead the current public org snapshot by GitHub stars, giving a quick view of the projects the community has returned to most.
dgm-eval
211 starsEvaluation code for deep generative models, including analysis of generative model metrics and diffusion-model treatment.
Open repositoryDropoutNet
184 starsNeurIPS 2017 code for addressing cold start in recommender systems, including training and evaluation routines.
Open repositoryxpool
138 starsProject code and site for cross-modal language-video retrieval research.
Open repositoryTabDPT-inference
97 starsLightweight inference code for TabDPT, an open source tabular foundation model trained on real-world data.
Open repositoryT-Fixup
90 starsICML 2020 code for improving transformer optimization through better initialization.
Open repositoryRecSys2018
88 starsLayer 6's first-place approach to the 2018 ACM RecSys Challenge.
Open repositoryRecent activity
Current releases show an active applied research mix.
The latest public pushes cluster around tabular foundation models, causal inference, agent evaluation, representation learning, and generative modeling.
forest-diffusion-mo
Diffusion and flow matching with XGBoost models for calorimeter data.
CCPFN-inference
Inference code for causal foundation models with continuous treatments.
TabDPT-inference
Tabular foundation model inference code for new prediction tasks.
Agentic-Monte-Carlo
ICML 2026 code for simulating reinforcement learning for black-box agents.
Community, publications, and project pages
Open source activity connected to research communities.
Several projects include dedicated pages, paper links, or conference-linked repository descriptions so readers can move between code, artifacts, and research context.
Open source at Layer 6
Explore, reproduce, and extend the research.
The full organization remains the source of truth for repository metadata, licenses, issues, and the newest releases.
Visit layer6ai-labs on GitHub