AI at Meta

company

Verified

https://ai.facebook.com/

facebookresearch

AI & ML interests

None defined yet.

Recent Activity

ccsasuke published a dataset about 6 hours ago

ccsasuke updated a dataset about 6 hours ago

ccsasuke submitted a paper about 8 hours ago

Two-Level Meta-Rubrics for Evaluating Open-Ended Generation: GAMUT, a Benchmark for Factual Completeness

View all activity

Papers

Two-Level Meta-Rubrics for Evaluating Open-Ended Generation: GAMUT, a Benchmark for Factual Completeness

Sparse Delta Memory: Scaling the State of Linear RNNs through Sparsity

View all Papers

facebook 's papers 68

Submitted by

Xilun Chen

Two-Level Meta-Rubrics for Evaluating Open-Ended Generation: GAMUT, a Benchmark for Factual Completeness

Submitted by

Niels Rogge

Sparse Delta Memory: Scaling the State of Linear RNNs through Sparsity

Submitted by

taesiri

TUA-Bench: A Benchmark for General-Purpose Terminal-Use Agents

Submitted by

taesiri

Autodata: An agentic data scientist to create high quality synthetic data

Submitted by

Vijay Viswanathan

Discretizing Reward Models

Submitted by

Nicolas Beltran-Velez

The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RL

2

Submitted by

Xichen Pan

RepFusion: Leveraging Multimodal Priors for Denoising in Representation Space

2

Submitted by

Jiasenlu

MobileMoE: Scaling On-Device Mixture of Experts

Submitted by

Rawal Khirodkar

Sapiens2

Submitted by

Niels Rogge

Scaling Test-Time Compute for Agentic Coding

Submitted by

Lei Zhang

Think in Strokes, Not Pixels: Process-Driven Image Generation via Interleaved Reasoning

Submitted by

Yuhang Zhou

Synthetic Sandbox for Training Machine Learning Engineering Agents

2

Submitted by

Daniel Gilo

Realiz3D: 3D Generation Made Photorealistic via Domain-Aware Learning

Submitted by

Stan Szymanowicz

LagerNVS: Latent Geometry for Fully Neural Real-time Novel View Synthesis

Submitted by

taesiri

Reasoning over mathematical objects: on-policy reward modeling and test time aggregation

Omnilingual SONAR: Cross-Lingual and Cross-Modal Sentence Embeddings Bridging Massively Multilingual Text and Speech

Submitted by

Niels Rogge

Omnilingual MT: Machine Translation for 1,600 Languages

5

Submitted by

Yixin Liu

Examining Reasoning LLMs-as-Judges in Non-Verifiable LLM Post-Training

Submitted by

taesiri

Multi-Task Reinforcement Learning for Enhanced Multimodal LLM-as-a-Judge

A recipe for scalable attention-based MLIPs: unlocking long-range accuracy with all-to-all node attention

Submitted by

Peter Tong

Beyond Language Modeling: An Exploration of Multimodal Pretraining

6

Submitted by

Xiaoke Huang

VecGlypher: Unified Vector Glyph Generation with Language Models

Submitted by

Longfei Yun

Decoding ML Decision: An Agentic Reasoning Framework for Large-Scale Ranking System

Submitted by

KAIQU LIANG

Learning Personalized Agents from Human Feedback

Submitted by

Qiang

Principled Synthetic Data Enables the First Scaling Laws for LLMs in Recommendation

Submitted by

Bhavul Gauri

AIRS-Bench: a Suite of Tasks for Frontier AI Research Science Agents

Submitted by

Julia Kempe

Teaching Models to Teach Themselves: Reasoning at the Edge of Learnability

3

Submitted by

Rémy Sabathier

ActionMesh: Animated 3D Mesh Generation with Temporal 3D Diffusion

Submitted by

Jay

Memorization Dynamics in Knowledge Distillation for Language Models

2

Submitted by

Yawar Nihal

ShapeR: Robust Conditional 3D Shape Generation from Casual Captures

Submitted by

taesiri

Inference-time Physics Alignment of Video Generative Models with Latent World Models

5

Submitted by

taesiri

VideoAuto-R1: Video Auto Reasoning via Thinking Once, Answering Twice

Submitted by

Yuanhao Cai

PhyGDPO: Physics-Aware Groupwise Direct Preference Optimization for Physically Consistent Text-to-Video Generation

Submitted by

Shashwat Goel

Training AI Co-Scientists Using Rubric Rewards

Submitted by

Haonan Qiu

HiStream: Efficient High-Resolution Video Generation via Redundancy-Eliminated Streaming

Submitted by

taesiri

SAM Audio: Segment Anything in Audio

Submitted by

Zeyuan Allen-Zhu

Physics of Language Models: Part 4.1, Architecture Design and the Magic of Canon Layers

Submitted by

Yushi Hu

Multimodal RewardBench 2: Evaluating Omni Reward Models for Interleaved Text and Image

Submitted by

Tariq Berrada

Flowception: Temporally Expansive Flow Matching for Video Generation

Submitted by

Zhaochong An

OneStory: Coherent Multi-Shot Video Generation with Adaptive Memory

2

Submitted by

taesiri

Scaling Zero-Shot Reference-to-Video Generation

Submitted by

taesiri

TV2TV: A Unified Framework for Interleaved Language and Video Generation

Submitted by

taesiri

TUNA: Taming Unified Visual Representations for Native Unified Multimodal Models

Submitted by

taesiri

WorldGen: From Text to Traversable and Interactive 3D Worlds

Submitted by

taesiri

SAM 3: Segment Anything with Concepts

Submitted by

taesiri

SAM 3D: 3Dfy Anything in Images

Submitted by

Amar Budhiraja

What Does It Take to Be a Good AI Research Agent? Studying the Role of Ideation Diversity

3

Submitted by

Amar Budhiraja

Souper-Model: How Simple Arithmetic Unlocks State-of-the-Art LLM Performance

Submitted by

L

Mixture of States: Routing Token-Level Dynamics for Multimodal Generation

Submitted by

Hanqing Zhu

The Path Not Taken: RLVR Provably Learns Off the Principals

Submitted by

taesiri

DigiData: Training and Evaluating General-Purpose Mobile Control Agents

Submitted by

taesiri

CRAG-MM: Multi-modal Multi-turn Comprehensive RAG Benchmark

Submitted by

Jason Weston

SPICE: Self-Play In Corpus Environments Improves Reasoning

Submitted by

Deren Lei

Beyond Reasoning Gains: Mitigating General Capabilities Forgetting in Large Reasoning Models

1

Submitted by

Niels Rogge

The Art of Scaling Reinforcement Learning Compute for LLMs

Submitted by

Hritik Bansal

HoneyBee: Data Recipes for Vision-Language Reasoners

Submitted by

Jack Zhang

The Alignment Waltz: Jointly Training Agents to Collaborate for Safety

2

Submitted by

Lin Long

Hybrid Reinforcement: When Reward Is Sparse, It's Better to Be Dense

4

Submitted by

Niels Rogge

OneFlow: Concurrent Mixed-Modal and Interleaved Generation with Edit Flows

4

Submitted by

Jacob Kahn

CWM: An Open-Weights LLM for Research on Code Generation with World Models

Submitted by

Zhepei Wei

TruthRL: Incentivizing Truthful LLMs via Reinforcement Learning

Submitted by

zhipeng cai

DepthLM: Metric Depth From Vision Language Models

Submitted by

Chuanyang Jin

The Era of Real-World Human Interaction: RL from User Conversations

3

Submitted by

Tianjian Li

Jointly Reinforcing Diversity and Quality in Language Model Generations

Submitted by

Niels Rogge

DINOv3

Hearing Anywhere in Any Environment

Submitted by

Jianyuan Wang

VGGT: Visual Geometry Grounded Transformer

Submitted by

Niels Rogge

Cluster and Predict Latents Patches for Improved Masked Image Modeling