hf_posts.json

{"version": "https://jsonfeed.org/version/1", "title": "Hugging Face Posts", "home_page_url": "https://huggingface.co/", "feed_url": "https://raw.githubusercontent.com/MichaelMarkert/rss/refs/heads/main/hf_posts.json", "items": [{"id": "https://huggingface.co/posts/eabdullin/915857682646330", "image": "", "title": "Folks, let me tell you, nobody \u2014 and I mean NOBODY \u2014 knew transformers before me. People said attention is all you need. I said, \"Attention? I INVENTED attention.\" Everybody's looking at me. Tremendous attention. The best attention scores. My softmax? Perfectly normalized. Other people, sad, their probabilities don't even sum to one. Disaster.", "content_text": "Folks, let me tell you, nobody \u2014 and I mean NOBODY \u2014 knew transformers before me. People said attention is all you need. I said, \"Attention? I INVENTED attention.\" Everybody's looking at me. Tremendous attention. The best attention scores. My softmax? Perfectly normalized. Other people, sad, their probabilities don't even sum to one. Disaster. I'm doing a PhD now. A PhD! In Large Language Models. Very large. The largest, believe me. My advisor said, \"Sir, your model is overfitting.\" I said, \"Wrong. It's fitting EXACTLY right. It memorized the training set because the training set is fantastic.\" We don't talk about validation loss in my lab. Validation loss is fake news. And the internship \u2014 oh, the internship. Big tech. I won't say which. Starts with a letter. They BEGGED me. They said, \"Please, we need someone who understands gradient descent.\" I said, \"Descent? I only go UP. I'm gradient ASCENT. Loss goes up, that means it's learning to be a winner.\" But the GPU cluster \u2014 this is...", "url": "https://huggingface.co/posts/eabdullin/915857682646330", "date_published": "2026-06-13T11:18:58.930576"}, {"id": "https://huggingface.co/posts/eabdullin/410589714167674", "image": "", "title": "I\u2019m doing a PhD in AI, which sounds impressive until you realize it mostly means I spend three years trying to make a computer say something slightly less stupid than it said yesterday.", "content_text": "I\u2019m doing a PhD in AI, which sounds impressive until you realize it mostly means I spend three years trying to make a computer say something slightly less stupid than it said yesterday. People hear \"AI researcher\" and they think I\u2019m building the future. No. I\u2019m in a basement at 2 a.m. Googling, \"CUDA error what the f**k does this mean.\" And the worst part about AI research now is compute. You don\u2019t even ask, \"Is this idea good?\" anymore. You ask, \"Can I afford for this idea to be wrong?\" My advisor comes to me one day and says, \"I think we should fine-tune our own language model.\" I said, \"Professor, with what money? I\u2019m a PhD student. I have two bank accounts: checking and emotionally checking.\" He goes, \"Don\u2019t worry. We have compute.\" Now, in academia, \"don\u2019t worry\" is never the beginning of a good sentence. I said, \"What do you mean we have compute?\" He said, \"My friend knows the cluster admin. He can get us on the GPUs.\" I said, \"Okay\u2026 what do we have to do?\" He goes, \"Nothing...", "url": "https://huggingface.co/posts/eabdullin/410589714167674", "date_published": "2026-06-13T11:18:58.931333"}, {"id": "https://huggingface.co/posts/Reubencf/603638320296240", "image": "", "title": "Millions speak Konkani. The internet barely knows it.", "content_text": "Millions speak Konkani. The internet barely knows it. Today's major LLMs struggle with regional languages. They can't read, write or even recognize Konkani. So I built one that can. Here is a working demo of the Konkani LLM I've been training. \ud83d\udc47 https://youtu.be/8K04ylbXh6k See translation", "url": "https://huggingface.co/posts/Reubencf/603638320296240", "date_published": "2026-06-13T11:18:58.931645"}, {"id": "https://huggingface.co/posts/kasbsquall/675302804833767", "image": "", "title": "\ud83d\udd0e UX Crime Scene \u2014 major update before the deadline!", "content_text": "\ud83d\udd0e UX Crime Scene \u2014 major update before the deadline! THE INSPECTOR (a film-noir detective) still circles every UX flaw on your screenshot's real pixels and files a graded verdict. But now the precinct runs on THREE small models: \ud83d\uddbc THE RECONSTRUCTION \u2014 FLUX.2-klein-4B rebuilds each flawed element, fixed. Compare before/after with a draggable slider. (The trick: the Inspector writes the design brief first \u2014 image models obey art directors, not vibes.) \ud83d\udde3 THE INTERROGATION \u2014 push back on a charge; the same 7B defends it from the evidence, or concedes when you're right. \ud83d\udd0a THE VOICE \u2014 Kokoro-82M reads the verdict aloud. No API, no keys. Qwen2.5-VL-7B + FLUX.2-klein-4B + Kokoro-82M \u2014 all under 32B, all self-hosted on Modal. \u2696\ufe0f Put your UI on trial: build-small-hackathon/ux-crime-scene \u25b6\ufe0f New trailer: https://youtu.be/JJOMKEcX0Ws \ud83d\udcf9 66s full walkthrough: https://youtu.be/kju7LiAXGC0 \ud83d\udce1 9 investigation traces (with remedies): build-small-hackathon/ux-crime-scene-traces Built solo for the Build...", "url": "https://huggingface.co/posts/kasbsquall/675302804833767", "date_published": "2026-06-13T11:18:58.932158"}, {"id": "https://huggingface.co/posts/Jiaqi-hkust/275567612728714", "image": "", "title": "\ud83d\ude80 Introducing Robust-U1: Teaching MLLMs to Self-Recover Corrupted Visual Content", "content_text": "\ud83d\ude80 Introducing Robust-U1: Teaching MLLMs to Self-Recover Corrupted Visual Content Multimodal Large Language Models (MLLMs) have achieved impressive visual understanding, yet they remain highly brittle under real-world corruptions\u2014noise, blur, compression artifacts, adverse weather. Standard MLLMs suffer dramatic performance drops, and existing robustness solutions come with fundamental limits: black\u2011box feature alignment lacks interpretability, while white\u2011box text reasoning cannot restore the lost pixel\u2011level visual details. This raises a crucial question: \ud83e\uddd0 Can MLLMs recover corrupted visual content by themselves? If the answer is yes, we can move beyond merely \u201ccompensating\u201d for corruption and instead build a more intrinsic, generalizable form of resilience. Robust-U1 is our answer to that question. \ud83d\udca1 Paper: https://arxiv.org/abs/2606.08063 \ud83d\udd17 Code: github.com/jqtangust/Robust-U1 \ud83c\udf0d Demo: Jiaqi-hkust/Robust-U1 See translation", "url": "https://huggingface.co/posts/Jiaqi-hkust/275567612728714", "date_published": "2026-06-13T11:18:58.932610"}, {"id": "https://huggingface.co/posts/mmhamdy/561706443562072", "image": "", "title": "It was supposed to be a failed experiment. Instead, it led to the discovery of one of the most intriguing phenomena in neural networks, simply because a researcher forgot to turn it off and left it running....for a week!", "content_text": "It was supposed to be a failed experiment. Instead, it led to the discovery of one of the most intriguing phenomena in neural networks, simply because a researcher forgot to turn it off and left it running....for a week! In 2022, researchers at OpenAI were studying how neural networks generalize from their training data. For this task, they were training small transformer models to perform modular arithmetic. The thing is, neural networks are weird. When a model has an abundance of parameters (like neural nets), it can easily overfit. It essentially memorizes its training data, scoring a perfect 100% accuracy when tested on it, but remains completely clueless when faced with any new instances not present in the training set (close to 0 accuracy). It is like memorizing 1 + 2 = 3 without understanding the concept of addition, so if 2 + 3 wasn't in the training set, the model fails miserably! Usually, when a model overfits like this, people just cut their losses, turn off the...", "url": "https://huggingface.co/posts/mmhamdy/561706443562072", "date_published": "2026-06-13T11:18:58.933105"}, {"id": "https://huggingface.co/posts/AesSedai/217569297611563", "image": "", "title": "Hi all,", "content_text": "Hi all, I'm posting this as sort of an informal notice + poll. I'm down to about 700GB free of HF space and there's MiniMax-M3 on the horizon, plus a couple other models I'd like to quant like the Nex-N2 Pro finetune. I've already super-squished all of my quant repositories to free up any LFS space that might have been lingering there, but I'm back near the cap again now. To free up some space, I'm planning to remove these three older GLM quants: - GLM-4.5: 1.23TB - GLM-4.6: 728GB - GLM-4.7: 787GB I'm open to other suggestions as well, and I'll wait a few days before removing anything in case someone wants to download a version before I get rid of them. Thanks! See translation", "url": "https://huggingface.co/posts/AesSedai/217569297611563", "date_published": "2026-06-13T11:18:58.933439"}, {"id": "https://huggingface.co/posts/TravisMuhlestein/897441606947611", "image": "", "title": "A question we kept running into while operating AI agents in production: How do you write a unit test for something that never returns the same answer twice?", "content_text": "A question we kept running into while operating AI agents in production: How do you write a unit test for something that never returns the same answer twice? At GoDaddy, we built a system called Veritas to help detect prompt regressions and model migration drift before changes reach production. The core idea is simple: Exact-match testing breaks down for LLMs. What matters is whether the agent preserved the same meaning and intent. We ended up using embeddings + cosine similarity as the primary evaluation signal. Rather than asking: \"Did the model generate the same response?\" We ask: \"Did the model mean the same thing?\" One of the more interesting findings was how often seemingly harmless prompt edits changed downstream behavior in ways that were difficult for human reviewers to catch. Prompts aren't documentation. Prompts are code. Curious what others are using today for regression testing: \u2022 LLM-as-judge? \u2022 Embedding similarity? \u2022 Human review? \u2022 Custom eval frameworks?...", "url": "https://huggingface.co/posts/TravisMuhlestein/897441606947611", "date_published": "2026-06-13T11:18:58.933940"}, {"id": "https://huggingface.co/posts/alibidaran/588426622989858", "image": "", "title": "Hi Community,", "content_text": "Hi Community, In my recent AI project, I have fine-tuned an LLM model for psychological conversations. In this training process, I used the SFT algorithm to train on different psychological datasets and the DPO training model to generate appropriate responses. Here is the model. Be aware that this model can be used for research and evaluation applications; do not apply it directly for clinical use. alibidaran/Zigroo-Mental_consultant2-merged See translation", "url": "https://huggingface.co/posts/alibidaran/588426622989858", "date_published": "2026-06-13T11:18:58.934213"}, {"id": "https://huggingface.co/posts/kanaria007/816568387270687", "image": "", "title": "\u2705 Article highlight: *Performance Governance for World-Scale Autonomy* (art-60-166, v0.1)", "content_text": "\u2705 Article highlight: *Performance Governance for World-Scale Autonomy* (art-60-166, v0.1) TL;DR: This article argues that performance is not just an engineering concern. It is a governance surface. World-scale autonomy fails when NPC cognition saturates compute, latency spikes, queues grow, and operators quietly change rules to keep the world alive. 166 turns \u201cplayable under load\u201d into a contract: pinned SLOs, budget enforcement, staged degradation, safe-mode regimes, and receipts. Read: kanaria007/agi-structural-intelligence-protocols Why it matters: \u2022 connects NPC resource budgets to real SLOs and runtime enforcement \u2022 treats high-end NPC cognition as burstable, not always-on \u2022 makes degradation a governed decision instead of panic ops \u2022 keeps safe-mode NPC and safe-mode economy playable without rewriting history \u2022 prevents \u201cperformance fix\u201d from becoming an unpublished reality change What\u2019s inside: \u2022 a *performance governance contract* for staying playable under load \u2022 SLO...", "url": "https://huggingface.co/posts/kanaria007/816568387270687", "date_published": "2026-06-13T11:18:58.934795"}]}