Tag: #ai-papers
Writing on GPUs, LLMs, MLOps, Kubernetes — and mindset · 79 posts
SOTA Real-Time Video Analysis — Tracking, Understanding, Efficient Inference
A survey of SOTA trends in real-time video analysis. We cover video understanding tasks like action recognition, object tracking, and temporal segmentation, the SAM 2 family concept of video segmentation and tracking, tr
2026-06-30 · 21 min read #ai-papers#video-understanding#object-tracking#sam2#action-recognitionSOTA Speech Recognition and Synthesis — From Whisper to Codec Language Models
We survey recent trends in speech recognition (ASR) and synthesis (TTS). From HMM to CTC/attention, Whisper large-scale weak supervision, and from Tacotron to neural vocoders and codec language models, we trace the linea
2026-06-30 · 15 min read #ai-papers#speech-recognition#text-to-speech#whisper#neural-codecSOTA Autonomous Driving Perception — BEV, Occupancy, End-to-End
We survey recent trends in autonomous driving perception. From the perception-prediction-planning-control stack, BEV multi-camera fusion, occupancy networks 3D occupancy representation, vision-centric vs LiDAR fusion, to
2026-06-30 · 16 min read #ai-papers#autonomous-driving#bev#occupancy-network#end-to-endRobot Safety and Alignment — Trusting Powerful Robots
How can we trust increasingly powerful robots. We take a balanced look at physical safety, constrained reinforcement learning and safety layers, handling distribution shift, human-robot collaboration safety, verification
2026-06-29 · 17 min read #ai-papers#robotics#safety#alignment#reinforcement-learningRobots That Learn from Human Video — The Dream of Web-Scale Data
Can a robot learn from videos of people handling objects. We cover affordances and trajectories, the domain gap, representation learning and pre-training, one-shot imitation, web-video scale-up, and combining with robot
2026-06-29 · 19 min read #ai-papers#robotics#imitation-learning#representation-learning#videoRobot Foundation Models — One Policy for Many Jobs
We lay out the trend of robot foundation models that aim to do many jobs across many robots with a single policy. We cover the concept of a generalist policy, large-scale robot data such as Open X-Embodiment, cross-embod
2026-06-29 · 19 min read #ai-papers#robotics#foundation-model#generalist-policy#open-x-embodimentWorld Models — When Robots Imagine the Future
An explanation of the concept of world models (learning environment dynamics), covering model-based reinforcement learning, latent-space prediction, the robotic application of video prediction and generative models, plan
2026-06-29 · 17 min read #ai-papers#robotics#world-models#model-based-rl#planningSim-to-Real — Bringing What Was Learned in Simulation into Reality
An analysis of why robot policies learned in simulation collapse in reality (the reality gap), covering countermeasures such as domain randomization, domain adaptation, system identification, and digital twins, along wit
2026-06-29 · 17 min read #ai-papers#robotics#sim-to-real#domain-randomization#digital-twinHow Robots Learn — Imitation Learning and Reinforcement Learning
An overview of the four ways robots acquire skills, followed by a deep look at imitation learning (teleoperation, behavioral cloning, DAgger) and reinforcement learning (rewards, policies, exploration): their principles,
2026-06-29 · 16 min read #ai-papers#robotics#imitation-learning#reinforcement-learning#robot-learningThe Robots Eye — 3D Perception and SLAM
How robots see the world. We walk through RGB-D, LiDAR, and stereo sensors, the SLAM pipeline, point clouds and voxels, 6D pose estimation, deep-learning perception, and finally NeRF and 3D Gaussian Splatting, with code
2026-06-29 · 18 min read #ai-papers#robotics#slam#3d-perception#pointcloudHumanoid Whole-Body Control — Walking on Two Legs and Handling with Two Hands
From walking on two legs to handling objects with hands, we lay out the core ideas of bipedal locomotion and whole-body control. We cover ZMP and MPC, reinforcement-learning locomotion, balance and fall recovery, the int
2026-06-29 · 21 min read #ai-papers#robotics#humanoid#locomotion#whole-body-controlTactile Sensing and Dexterous Manipulation — Robots That Feel with Their Fingertips
Robots have begun to feel the world with their fingertips. We lay out vision-based tactile sensors (the GelSight family) and electronic skin, the role of touch in sensing slip, force, and texture, visual-tactile fusion,
2026-06-29 · 21 min read #ai-papers#robotics#tactile-sensing#manipulation#gelsightTwo Brains for a Humanoid — GR00T N1 and Helix
A VLA for humanoid robots must combine fast reflexes with slow deliberation. Centered on NVIDIA GR00T N1 and Figure AI Helix, we organize the dual-system architecture that combines fast low-level control (System 1) with
2026-06-27 · 15 min read #ai-papers#robotics#humanoid#groot-n1#helixRobots That See, Hear, and Move — A Review of VLA Models RT-2 and OpenVLA
Vision-Language-Action (VLA) models take camera images and natural-language instructions and output robot actions directly. Centered on RT-2, Open X-Embodiment, and OpenVLA, this post organizes the VLA paradigm: its idea
2026-06-27 · 15 min read #ai-papers#robotics#vla#rt-2#openvlaDiffusion Policy and π0 — The Secret Behind Smooth Robot Behavior
Going beyond the limits of discrete action tokens, we examine two streams that generate actions as continuous values. Diffusion Policy generates actions via denoising, and π0 produces high-frequency continuous actions wi
2026-06-27 · 14 min read #ai-papers#robotics#diffusion-policy#pi0#flow-matchingMultimodal AI Training Methods — Many Senses in One Model
A walkthrough of how multimodal AI learns to handle images, text, audio, and video in a single model. We cover modality alignment and contrastive learning, fusion strategies, shared embedding spaces, pretraining and fine
2026-06-26 · 15 min read #ai-papers#multimodal#clip#vision-language#contrastive-learningBeyond OCR — OCR-free Document Understanding and Unified Models
A traditional OCR pipeline splits into detection, recognition, and layout stages, but errors accumulate. We organize the shift in document AI: Donut-style and VLM-based OCR-free document understanding, high-resolution an
2026-06-26 · 20 min read #ai-papers#ocr-free#document-understanding#multimodal#donutAI Hardware Research Trends 2026 — The Future Through the Papers
A field-by-field review of where AI hardware research is heading in 2026. From wafer-scale and photonics, compute-in-memory, FP4 low-precision training, sparsity and MoE hardware, optical interconnect, next-generation me
2026-06-16 · 19 min read #ai-papers#ai-hardware#photonics#compute-in-memory#low-precisionLLM Landmark Papers Roundup 2026 - Transformer / Scaling Laws / Flash Attention / Mamba / DeepSeek-R1 / Titans Deep Dive
From the 2017 Attention Is All You Need paper to 2026 Titans and DeepSeek-R1, a thematic roundup of the 50+ landmark papers that built the LLM era. Transformer, BERT, GPT 1-3, Scaling Laws, Chinchilla, InstructGPT, PaLM,
2026-05-16 · 21 min read #llm-papers#transformer#scaling-laws#flash-attention#mambaMixture of Experts (MoE) Architecture Paper Deep Analysis: From GShard to DeepSeek-MoE
Analyzes core papers on Mixture of Experts architecture, comparing routing strategies and training stability techniques across GShard, Switch Transformer, Mixtral, and DeepSeek-MoE.
2026-03-14 · 34 min read #ai-papers#moe#transformer#deepseek