Blog post - Autoregression vs Diffusion - Understanding Sampling... via Optimal Transport Autoregression and diffusion look like opposites. Yet, beneath the implementation details, both are solving the same transport problem: how to turn simple noise into structured data.show more

In flow matching, a coupling determines how noise and... data samples are paired during training. The choice of coupling is important because it influences the geometry of trajectories at inference time. The simplest choice is the independent coupling, where noise and data points are paired arbitrarily. This can lead to curved trajectories as the model averages over many conflicting pairings. However, if we use optimal transport on batches of pairs, this leads to fewer ambiguous intersections that the model must resolve, leading to straighter trajectories at inference time.show more

Turn complex docs into clean, LLM-ready data! Every AI... company I've talked to is solving the same problem: how do you build systems that don't hallucinate and back up every answer with proper citations? Tensorlake is a tool that extracts custom-defined structured data from any unstructured document in 3 steps: ↳ Define your schema ↳ Enable citations ↳ Extract You get RAG-ready data with precise citations and bounding boxes. Feed this to your LLM, and you'll generate responses that are citation-backed and fully auditable. This is the difference between a demo and a production system. When your AI can show exactly where it got its information, you move from proof-of-concept to something people can actually trust and deploy. I've shared the Tensorlake GitHub repo in the replies!show more

Vision AI agents are becoming a practical way to... automatically turn video data from the physical world into operational intelligence in factories, cities, warehouses and transportation systems. See how Roboflow, DeepHow, and Linker Vision use NVIDIA Cosmos and NVIDIA Metropolis to generate synthetic data, fine-tune vision models faster, and deploy vision agents at scale. 📰 Read the blog:show more

Migrants are that relaxed that they actually use public... transport like local buses to get to the beaches. Does this look like they’re being stopped, pushed back, or anyone being prevented from reaching the beaches? It’s just a well-organised operation to bring them into our country. What exactly are we paying millions and millions of UK taxpayers’ money for?show more

Driven by a UK Government push to open up... transport data and boost public transport use nationwide, Google Maps has officially integrated live public transport tracking. By mandating that operators share their live GPS data, tech apps can now let you literally watch your specific bus moving along the map in real-time. As seen in the screenshot above, you can tap on your route (like the 207 towards White City) and see exactly where the bus is on the road, how many minutes away it is, and when the data was last updated. No more relying on broken countdown boards. No more running for a “ghost bus” that never actually shows up. No more standing in the rain guessing if you have time to grab a coffee. Have you noticed this government-backed feature on your phone yet? And more importantly... which West London bus route is the most unreliable and needs this the most? Let us know below #UB1UB2 #London #Southall #Hayes #Ealingshow more

Massive update for AI Engineers! Training diffusion models just... got a lot easier. dLLM is an open-source library that does for diffusion models what Hugging Face did for transformers. Here's why this matters: Traditional autoregressive models generate text left-to-right, one token at a time. Diffusion models work differently - they refine the entire sequence iteratively, giving you better control over generation quality and more flexible editing capabilities. The problem? Building and training these models required stitching together scattered tools and reimplementing research papers from scratch. Most of the tooling has been scattered and is hard to reproduce. dLLM changes this: It unifies everything you need to train, evaluate, and deploy diffusion language models: ↳ Scalable training with LoRA, DeepSpeed, and FSDP support ↳ Unified evaluation that abstracts away inference complexity ↳ Ready-to-use recipes for pretraining, finetuning, and evaluation The library includes implementations of models like LLaDA and Dream, plus training algorithms like Edit Flows that enable insertion, deletion, and substitution operations. The team just released ModernBERT-Chat models showing you can turn BERT into lightweight chatbots through masked instruction tuning. This is practical and worth exploring. The setup is straightforward. Run locally with Accelerate or scale to multi-node clusters with Slurm. If you're working with language models and want to explore diffusion-based approaches without rebuilding infrastructure, dLLM gives you a production-ready starting point. Link to the repo in the next tweet.show more

Car-to-curb distance estimation using Ultralytics YOLO26. 🚗 In parking... lots, knowing the exact distance between a vehicle and the curb can help prevent minor collisions and improve parking precision. In this demo, a car is detected and tracked, and the system calculates the distance from the curb in real time in pixel space, showcasing how computer vision can turn simple camera feeds into spatial awareness tools. Implementation details 👇 #cardistance #visionai #ROADSAFETYCONCERNshow more

We’ve seen humanoid robots walk around for a while,... but when will they actually help with useful tasks in daily life? The challenge here is the diversity and complexity of real-world scenes. Our new work tackles this problem via 3D visuomotor policy learning. Using data from only 1 scene, our Improved 3D Diffusion Policy (iDP3) enables a full-sized humanoid robot to autonomously pick&place objects, pour water, and wipe tables, in the wild open world. (and all these skills are useful, right?) Web: Fully open-sourced code:show more

💡Data Scraping vs. Data Mining: Understanding the Difference Ever... wondered why OptimAI Network emphasizes Data Mining over simple web scraping? Here’s why: 🔹Web Scraping is surface-level, capturing raw data from websites without context, validation, or depth. 🔹Data Mining, however, is a deeper, intelligent process—extracting, analyzing, validating, and refining data into structured insights essential for advanced AI models. 💫 Why OptimAI Focuses on Data Mining? OptimAI's decentralized nodes don’t just collect data—they actively validate, annotate, and refine it. By leveraging collective human intelligence, edge computing, and autonomous AI agents, we deliver the high-quality, real-world data necessary for powering sophisticated AI. 💫 Why It Matters—and Why You Should Join Now Participate now via our OptimAI Lite Node, and become a foundational part of building the most comprehensive decentralized Reinforcement Data Network for Agentic AI. Mine data, fuel innovation, and earn OPI rewards. 👉 Chrome Extension Node: 👉 Telegram Node: Together, let's redefine data for #DePIN #AI.show more

Asking a fit person what they do to stay... fit AND assuming that’s how they got fit is like asking a rich person how they maintain wealth and assuming that’s how they built it. The maintenance protocol and the building protocol are not the same thing. If you model the current behavior without the foundation, you’re solving the wrong problem. And yeah, the build takes some time… But once you get there? Maintenance is easy. If you want to start from the right place without trial and error, DM me “TRAIN”.show more

Another Day, Another Safety Hazard: Is KSRTC Ignoring Basic... Maintenance? Is this the new standard of bus maintenance in Karnataka? This KSRTC bus appears to be moving unevenly on the road instead of running straight, raising serious concerns about its condition and maintenance. Public transport vehicles carry hundreds of passengers every day, and even a small mechanical issue can put many lives at risk. The poor condition of the bus clearly reflects a lack of proper inspection and maintenance. If such vehicles are allowed to operate on busy roads, they not only endanger passengers but also other road users. Safety should be the highest priority for any transport authority, yet incidents like this raise questions about whether regular checks are being conducted effectively. The Transport Department and the Transport Minister Ramalinga Reddy must take accountability and ensure that all buses are thoroughly inspected before being put into service. Preventive maintenance is far cheaper than dealing with accidents and loss of life. If an accident occurs due to negligence, who will take responsibility? The authorities cannot continue to ignore these warning signs while citizens risk their safety every day.Public transport should inspire confidence, not fear. Passengers deserve safe, reliable, and well-maintained buses, and the government must ensure that taxpayer money is being used to maintain these vehicles to the highest standards. #bangalore #bengaluru #ksrtc DK Shivakumar Siddaramaiah BJP Karnataka alok kumar DGP KARNATAKA ಕರ್ನಾಟಕ ರಾಜ್ಯ ಪೊಲೀಸ್ Karnataka State Police S. Lalitha ChristinMathewPhilipshow more

Most recent diffusion language model research (that I’ve seen)... seems to be using masking as the noising process. It looks like, however, most closed-source models (Google Gemini Diffusion and possibly Inception Labs’ Mercury) use a different noising process, where instead of masking tokens, they replace them with different tokens (either with a random token or a semantically similar token). I wondered how they were getting such high throughput with the latter noising process, since I believed that optimizing inference with KVCache approximation would be more difficult (for various reasons). I visualized this noising process with tiny-diffusion and compared it to normal unmasking, and was very surprised to see how fast the generation “settles” into a reasonable output, and then only slightly refines afterwards, requiring much fewer steps in total. Unmasking (where tokens are never remasked, the typical implementation) is inherently limited in generation speed by the fact that an increase in tokens decoded per step leads to more errors due to the mismatch between individual and marginal token probability distributions we sample from. The token replacement noising process seems to have a much different set of characteristics. Because we sample each token per step, every token makes “progress” towards the final output each iteration (in addition to *potentially* giving other tokens more information in future steps). Generally, masking has outperformed other noising processes, which is probably why most research focused on it (using smaller models). But the paper referred to in the retweet shows that random replacement as a noising process may scale better as model size increases. Big labs might have noticed these results much earlier (due to having drastically more training resources and being able to test larger models), which may explain the discrepancy in the choice of noising process. I’m gonna test this with larger models, since tiny-diffusion only has 10M parameters.show more

#support2earn Metabows is a social problem-solving project utilizing WEB3,... initiated by Satoshi and his beloved dog Bruno. Metabows issues BOW tokens and uses the value generated from them to give back to society, aiming to achieve a "well-being society for both humans and dogs." BOW tokens can be mined by participating in a new X to earn mechanism called support2earn. As more people around the world spread the philosophy of Metabows through support2earn, the value of BOW tokens will increase, and society will improve. Metabows progresses through blog stories linked to the project, so please read the blog and participate in support2earn. ↓ ▪︎=============▪︎ 【How to Participate】 From 4/14 23:00 to 4/15 23:00 1⃣ Hold DFT or XCR 2⃣ Like the post 3⃣ Follow Metabows🐕 ADVENTURES 📿🎖️ 4⃣ Quote the first tweet Add a comment ＋ Use any hashtag(s) (Please include easily spreadable hashtags, but it seems better not to use more than two) 5⃣ Participation pageshow more

FACT Check: Conservatives are attacking Biden saying he's requiring... Americans Trapped in Israel, who want to leave, sign a promissory note to repay costs? FACTS: This has been US policy for at least 79 years! Although the United States government is requiring individuals who intend to be evacuated sign a promissory note (known as the form DS-5528), this has been a US policy going back to at least 1944. President Trump required those evacuated from foreign nations sign the same note and reimburse costs just like Biden is and all the previous presidents have since 1944. Here's what the US State Department website (which was active during Trump's President says about Form DS-5528? "Generally, individuals evacuated on a U.S. government-coordinated transport, including charter and military flights or ships, even if those transports are provided by another country’s government, must sign an Evacuee Manifest and Promissory Note (Form DS-5528) note prior to departure. The Department of State uses the Form DS-5528 to document who got on which transport, and it lets us know how to contact evacuees for billing purposes. U.S. law requires that departure assistance to private U.S. citizens or third country nationals be provided “on a reimbursable basis to the maximum extent practicable.” By taking a U.S. government coordinated transport, evacuees are obligated to repay the cost of their transportation. The amount billed to evacuees is based on the cost of a full fare economy flight, or comparable alternate transportation, to the designated destination(s) that would have been charged immediately prior to the events giving rise to the evacuation. Because different evacuation transports may go to different destinations, the cost of one transport may be different than another. We encourage people to leave on the first transport they are able and eligible to board." FACTS MATTER. Stop spreading misinformation.show more

AI VIDEO NEURAL NETWORKS HAVE MASTERED URBAN DYNAMICS AND... COMPLEX DEPTH OF FIELD A vibrant redhead in a pink crop top and pleated mini skirt walks down the street as passersby turn to look. None of it was filmed in reality: not the fabric tension during motion, not the thin purse chain, nor the natural reactions of the background crowd. Rendering crowds and background objects is a major stress test for video diffusion. Secondary characters typically warp, while bicycle spokes and faces blur into digital noise. Here, street geometry, depth of field, and skirt pleats remain fully stable. Tracking a subject against a chaotic background used to break spatial coherence. Now, the model cleanly isolates layers: the focused subject, realistic urban parallax, and authentic crowd reactions without distortion. Generative video has advanced beyond simple studio setups to reliably handle dynamic street environments with live extras.show more

OAU transport system was near perfect, adequate buses and... bikes, we rarely experienced queues until the First Lady came and decided she didn’t like the ease of transportation and introduced “change” by donating CNG buses. The school management in turn banned all motorcycles and commercial buses on campus and subjected over 30,000 students to just about 50 buses and a few tricycles. OAU has one of the largest land masses of any Nigerian university, yet students are now stranded daily. This is a Monday morning where students are supposed to rush for class, but they are stuck at the campus gate because of the inefficiency of the CNG buses. Students are struggling to go to class in the morning and in the evenings to even get back to their hostels. Students staying off campus now spend almost double to get to campus, all alternative transport means have been removed The management have cars to bypass all of this, but majority of students do not. Yet they released statements talking about how the buses will “ease transportation” and even implied students appreciate the system. Meanwhile, students are trekking, missing classes, and spending hours in queues. Let it be clear that students do not appreciate this system. It has made our lives harder, not better. What used to be a smooth system has now become a daily struggle. There are many pressing issues in this school, but they chose to dismantle a working transport system and replace it with hardship. We can no longer remain silent. Let the whole world know what OAU students are going through. Please if you come across this tweet, help us retweet it whether you are an OAU student or not We demand our towngboro buses and motorcycles back.show more

Drivers like this are exactly how people end up... getting seriously hurt on the road. The cyclists clearly signalled a turn, yet this driver still chose to overtake straight into a developing hazard. Then a ute pulls out from the side street and suddenly you’ve got a near head-on. Completely avoidable if there was even a moment of patience and awareness.show more

Appreciation post for Minister Mhona Felix Tapiwa Mhona Yesterday,... I was kumusha ku Mhondoro. I decided to go and buy some building materials pa Mubaira. My driver had been telling me that the road that myself and other MPs from Mhondoro, including Mutsa Murombedzi and the current MP, Hon. Chamu Chiwanza, were making noise about in Parliament and Social Media was being fixed. As a responsible opposition MP, I must, in the same manner that I made noise about the state of the road, also come here today and appreciate the work done by the government through the Minister of Transport, Hon. Mhona, for listening to your fellow MPs and delivering. Even though it is still gravel, we must still appreciate it and say, "Thank you." Min Mhona , Hon Chamu Chiwanza the area MP and Myself we are ALL MHOFU 🤔show more