🚀Thrilled to share my first work as a PhD student! We propose a new home for Diffusion LLMs: not as competitors to AR models, but as ultra-fast drafters. DFlash is lightweight, cheap to run, and very effective (up to 6x speedup). It’s super easy to set up—give it a try!

𝗧𝗵𝗶𝗻𝗸 𝘄𝗶𝗱𝗲𝗿. 𝗧𝗵𝗶𝗻𝗸 𝘀𝗵𝗼𝗿𝘁𝗲𝗿. 🚀 𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗠𝘂𝗹𝘁𝗶𝗽𝗹𝗲𝘅 𝗧𝗵𝗶𝗻𝗸𝗶𝗻𝗴: token-wise branch-and-merge reasoning for LLMs. 💸 Discrete CoT is costly. 🎛️ Existing continuous tokens often clash with 𝗼𝗻-𝗽𝗼𝗹𝗶𝗰𝘆 𝗥𝗟

⚡️ Exciting news: DFlash is now on SGLang! We are unlocking new possibilities to accelerate LLM inference. 🚀 Stay tuned—more draft models and optimizations are dropping soon that will seriously speed up your workflows!

One of the best ways to reduce LLM latency is by fusing all computation and communication into a single GPU megakernel. But writing megakernels by hand is extremely hard. 🚀Introducing Mirage Persistent Kernel (MPK), a compiler that automatically transforms LLMs into optimized

🔥 We introduce Multiverse, a new generative modeling framework for adaptive and lossless parallel generation. 🚀 Multiverse is the first open-source non-AR model to achieve AIME24 and AIME25 scores of 54% and 46% 🌐 Website: https://t.co/J9osByhWUf 🧵 1/n

🥳 Happy to share our new work –  Kinetics: Rethinking Test-Time Scaling Laws 🤔How to effectively build a powerful reasoning agent? Existing compute-optimal scaling laws suggest 64K thinking tokens + 1.7B model > 32B model. But, It only shows half of the picture! 🚨 The O(N²)

🚨 Why Do Multi-Agent LLM Systems Fail? ⁉️ 🔥 Introducing MAST: The first multi-agent failure taxonomy - consists of 14 failure modes and 3 categories, generalizes for diverse multi-agent systems and tasks! Paper: https://t.co/BC5YHS8ZRZ Code: https://t.co/Ea1FvGcaLs 🧵1/n

🚨 Thrilled to present our Spotlight at #ICLR2025: "MagicPIG: LSH Sampling for Efficient LLM Generation" by @RJ_Sadhukhan 🎉 💡 MagicPIG enables KV compression for long-context LLMs — where top-k falls short, sampling shines. ⚙️ Introduces CPU-GPU heterogeneous serving to boost

🚀 Low latency, high throughput, and lossless LLM inference — all at once. 🎂 Who says you can’t have your cake and eat it too? #MagicDec proves you can. ⚡ MagicDec first shows that speculative decoding can boost throughput for moderate to long contexts — and the speedup grows

🚀 RAG vs. Long-Context LLMs: The Real Battle ⚔️ 🤯Turns out, simple-to-build RAG can match million-dollar long-context LLMs (LC LLMs) on most existing benchmarks. 🤡So, do we even need long-context models? YES. Because today’s benchmarks are flawed: ⛳ Too Simple –