HotOpen discussion
When a favorite model disappears, what exactly do we lose?
Beyond benchmark scores: style, reproducibility, research continuity and migration cost.
The research commons · EN / 中文
Trace the signal.
Build what matters.
Research news with context, papers explained without hand-waving, tools you can actually run, and a community that helps ideas travel across fields.
Watch
Spend attention before the opportunity or deadline passes.
Build
Turn a useful signal into an experiment, artifact or collaboration.
Since your review
–new signals
Creating a private local baseline…Stored only in this browser
Signal desk
What changed in research
Official sources first · refreshed daily by an automated build · editorial fallback when feeds fail · Scores are transparent editorial heuristics, not citation metrics
Deadline desk
Still in time to participate
Frontier watch
New builders shaping the next cycle
Model
World Labs · Atlas
It unifies controlled generation, sparse-view reconstruction and real-to-sim robotics inside one spatial context.
World Models◷ Deep dive
Lab
Renesas Physical AI & Robotics Lab
It connects embodied-model progress to chips, sensing, control and deployable robotics systems.
Embodied AI◷ Skim
Lab
NVIDIA–KAIST Joint AI Research Lab
It pairs academic research with frontier compute and a clear regional deployment agenda.
Agentic AI◷ Read
Community-tested
Research tools worth trying
Agent SkillsCurated
AREX-Skill + DisCo
The missing skill layer between intelligence and research
88Impact
94Buzz
93Utility
Paper SearchCurated
DeepXiv SDK
An agent-native CLI and index for all of arXiv
78Impact
82Buzz
91Utility
People SearchCurated
DINQ
Turn an underspecified people-search request into verifiable candidates
84Impact
87Buzz
86Utility
People in the loop
Ask, share, build together.
Research help, internships, collaborators and tool launches belong in one place — with a path from quick chat to durable public knowledge.
Enter community →QuestionOpen discussion
Show us your paper-reading workflow
Zotero, Obsidian, agents or pen and paper — share the system that survives a busy week.
BuildOpen discussion
Vote for the next interactive demo
KV cache calculator, MoE routing playground, or an eval-design clinic?
Counts are local interest signals from this browser only; no community totals are implied.
AI & CS · Original collection
Interactive paper deep-dives
LLaDA: Large Language Diffusion with mAsking
Nie et al. · arXiv 2025
A masked diffusion framework for LLMs. Uses progressive masking as the forward process and learns to predict masked tokens in reverse, matching AR models at 8B scale.
Diffusion LMMasked DiffusionPre-training
Simple and Effective Masked Diffusion Language Models (MDLM)
Sahoo et al. · NeurIPS 2024
Simplifies masked discrete diffusion with a principled continuous-time ELBO. Clean, minimal design with strong perplexity results.
Diffusion LMMasked DiffusionELBO
Fast Discrete Diffusion Language Models (Fast DLLM)
Shi et al. · arXiv 2025
Accelerates discrete diffusion LMs with adaptive noise schedules and importance sampling, reducing denoising steps by 3-10x.
Diffusion LMFast SamplingAdaptive
Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Models
Arriola et al. · arXiv 2025
Generates text in blocks — blocks go left-to-right (AR), tokens within each block are denoised in parallel (diffusion). Best of both worlds.
Diffusion LMBlock GenerationHybrid