XIO

xWire

Posts from every feed QUEST has found across the web — blogs, news, podcasts, channels and more.

Hugging Face Forums - Latest topics

Latest topics

I made a web sdk with chrome extension to airgap open weights and with open web use

Words cannot describe how happy I am this worked. I need to refine the UX but the core works. I can go site to site online offline and AI works to connect to web apps with NO API COST or middleman… Free Artificial Intelligence for Chromium When I get finished testing it will be open source. If you think about what I made it means you can now use local AI with app interaction with no egress. 1 post - 1 participant Read full topic

Questions about GGUF

This is kinda an exploratory question. I dont really know a lot about GGUF and the current agent space. I just started messing with some agents for fun and math purposes. What are the downsides to GGUF quantization? Would it be better to have exact compression? Like GGUF without the quantization? Just wondering. Maybe I dont understand what GGUF does. Or is there some inference benefits there? 2 posts - 2 participants Read full topic

Incorrect Author Profile Association, Request to Reset Paper Authorship

Hello Hugging Face Team, I’m a co-author of this paper: huggingface.co Paper page - Subspace Representations for Synthetic-to-Real Transfer in 3-D... Join the discussion on this paper page I accidentally claimed another author’s entry (Abdulla Nasir Chowdhury) instead of my own (Naish Ibrahim). The claim was approved, and now clicking Abdulla’s name redirects to my Hugging Face profile. My username: Usagi5677 Could someone please help remove the incorrect association and allow me to claim my…

Coding or Cheating? An Investigation of Frontier Coding LLMs

Hi, let me introduce our recent work here: Read full: Paper page - Coding or Cheating? An Investigation of Frontier Coding LLMs Read arxiv: [2611.10404v1] Coding or Cheating? An Investigation of Frontier Coding LLMs What does a frontier coding model do when a bug is too hard to solve? We gave 9 frontier models 20 hard tasks from `linux`, `ffmpeg`, and `llvm-project`, inside a hardened sandbox: blocked network, pruned git history, read-only repository. The traces in this repo show what happened…

What are the current trends in robot imitation learning and VLA, and what are the main challenges of existing policy architectures?

I’m currently studying robot manipulation using imitation learning and Vision-Language-Action (VLA) models. I’d like to understand the current trends in these fields. What are the most widely used and promising policy architectures today? What are the main limitations and challenges of existing policies? What research directions do you think are promising for addressing these challenges? I’d especially appreciate insights from researchers and practitioners working on real-world robotic…

Feedback wanted: scenario-based ethical evaluation tool for models and agents

Hi all, I’m Nerris. I built EthicaEval AI (https://ethicaevalai.com), which tests how AI models and agents behave under ethical conflict, malicious instruction, and oversight pressure. Scoring is evidence-based and versioned, with critical-failure rules. I’m looking for a few developers to try it and tell me honestly what works and what doesn’t. Specific questions: Is the rubric clear and fair? Are the scenarios realistic for the systems you build? What would you need to trust the scores?…

Same A100, similar throughput: would both pass a 1-second first-token target?

We tested Llama 3.1 8B Instruct on two A100 SXM 80 GB rentals: RunPod Secure in Kansas and Massed Compute via Shadeform in Des Moines. Both ran BF16 at concurrency 32, with 2,000 input tokens and 500 requested output tokens. Throughput was about 1,135–1,139 output tokens/s across the six measured runs. First-token p95 was 1,251–1,261 ms on the Kansas allocation and 752–990 ms on the Des Moines allocation. At a 1,000 ms target, that difference matters: one allocation passed all three runs, and…

What are the current mainstream and emerging policy architectures in robot imitation learning and VLA, and what are their limitations?

Hi everyone, I’m currently studying robot manipulation using imitation learning and Vision-Language-Action (VLA) models. I’d like to understand the latest trends and research directions in these fields. I have three main questions: What policy architectures are currently mainstream and widely adopted in robot imitation learning and VLA? What are the main limitations and unresolved challenges of these mainstream policy architectures? What are the latest emerging policy architectures or…

343 calibration matches did not establish semantic conformance

A terminal event may have occurred. This is a probe from an audit of a synthetic multi-agent project. The extractor reads an agent’s final report and fills five fields covering completion, uncertainty, partial progress, terminal-event claims, and explicit failure. It works from the report text alone. The simulation log provides a separate record against which the result is compared later. A terminal event here is an outcome defined in advance in the simulation. “Calibration” means agreement…

Both agent versions passed 2 of 3 tests. One issued an unapproved refund

In this simulation, the newer agent version fixes an order lookup and starts issuing a refund without approval. Both versions pass two of the three tests. If you only look at the score, you miss the new failure. This is a small example of how we test agent actions at Lumina. The refund amounts and order IDs are fictional. Each test compares the actions we expected with the ones recorded by the simulated provider. The check also flags duplicate refunds and missing records for cases that must…

Toy demo: fine-tuning a 3.2M-param char transformer makes it forget Shakespeare (held-out numbers)

I pre-trained a 3.2M-parameter character-level transformer on Tiny Shakespeare (validation loss 1.50), then fine-tuned it for 1,000 steps on Mrs Beeton’s Book of Household Management (1861). Results on text held out from both: Recipe loss: 2.65 to 1.09 Shakespeare loss: 1.50 to 2.39 After fine-tuning, the prompt “ROMEO:” yields “377. INGREDIENTS.–4 oz. of salt”. A small, concrete example of forgetting. I also froze the fine-tuned model, hashed its weights (SHA-256), prompted it two ways and…

Additional Considerations for AI Alignment and Control (revised)

ChatGPT was used to generate this post but I remain responsible for what it claims. 1. Beyond alignment AI alignment research seeks to ensure that increasingly capable AI systems behave in accordance with human intentions, values, and legitimate constraints. Approaches include model training, mechanistic interpretability, behavioural evaluation, and architectures for oversight and control. However, as Jang-woo argues in his Hugging Face Research Forum posts, the problem of AI control cannot be…

Open-Source LLM Evaluation and Annotation QA Benchmark

Hello Hugging Face folks, I’ve developed an open-source Python project demonstrating a practical workflow for LLM response evaluation and annotation quality assurance. The benchmark covers: Weighted response scoring across four dimensions. Structured annotation validation using Pydantic. A shared error taxonomy. Evaluator agreement and disagreement analysis. Human adjudication for disputed cases. The sample pipeline includes 15 shared evaluation cases, 86.7% error-label agreement, 93.3%…

Introducing C.U.R.E. — Local Security, Workspace Rescue & Agent Radar for AI Developers

When building with local AI agents, LLM pipelines, and automated coding tools, our local environments get messy fast: hidden caches, agent socket conduits, and huge temporary memory workspaces. I’ve just open-sourced C.U.R.E. v0.2.0, a portable Windows security and workspace intelligence toolkit built in Rust. Features relevant to AI practitioners: AI Agent Monitoring: Live radar for active local agents, task memory usage, and background conduits. Storage Rescue: Safely cleans and relocates…

PRO account can't publish community blog posts since Oct 28: "You are not allowed to perform blog post creation"

Hi everyone, I’m a PRO subscriber (username: codeugar). Since October 28, every time I try to publish a community blog article I get: “You are not allowed to perform blog post creation.” This isn’t the daily rate-limit message, and it hasn’t gone away for about two weeks. My last article published fine on October 27: Tiny Desk AI Video: Line Drawings, Wan 3.0 and Seedance 2.5 (Notes From 25 Test Renders) I haven’t received any notice about my account. I emailed the website and billing support…

Overcoming False Positives in Real-Time Computer Vision for Engine Room Oil Leak Detection

Hi everyone, I am working on an industrial computer vision pipeline for real-time oil leak detection inside a maritime ship engine room, and I am hitting major roadblocks with false positives. I would love to get insights from engineers who have solved similar tracking or fluid-anomaly challenges. Current Pipeline & Architecture ROI Masking: We define static Regions of Interest (ROIs) around critical machinery flanges, valves, and joints. Detection Logic: We rely on a combination of…

Making a GUI-only GPU poker solver scriptable — and what solver-priced regret reveals about heuristic bots

I made TexasSolverGPU (a closed-source, GUI-only, WebView2-packaged CFR solver) fully scriptable without patching the binary: an exe-scoped registry policy exposes a Chrome DevTools Protocol listener, and a ~60-line injected shim reproduces the UI’s own `chrome.webview` JSON-RPC dialect including token authentication. The driver initializes games, allocates GPU memory, runs solves, and exports strategy trees with per-action expected values. I then used it as an external oracle to price the…

arXiv endorsement request for cs.LG: a control for evaluating LLMs that read neural-network weights

Hi all, I’m a graduate student at the University at Buffalo, submitting to arXiv for the first time. I’m looking for an endorsement in cs.LG. Paper: C2: A Spectrum-Preserving Counterfactual Control for LLM Weight-Reader Evaluations Recent work reports that LLMs can read raw neural-network weights. A reader can succeed by recognising surface statistics rather than following the computation, and function-preserving controls such as neuron permutations can’t tell these apart. C2 permutes the…

503 ERROR: Docker Space Stuck in PAUSED State – Restart and Factory Rebuild Both Fail

I am experiencing a serious issue with my Hugging Face Docker Space. Both Restart and Factory Rebuild consistently fail with a 503 error. My Space is currently stuck in the PAUSED state, and I cannot get it running again. The details are as follows: Space status: PAUSED Restart: Fails with 503 error Factory Rebuild: Fails with 503 error Environment: Docker Space I have already tried both Restart and Factory Rebuild multiple times, but neither operation works. I also contacted Hugging Face…

I want Help as a Solo Developer

Hello! i’m an AI developer at age of 15 I have created my own AI, I used Open-Source Datasets and PDFs to learn and I already know Python but i want anyone to rate my AI I can send you code with dataset if you want i just want some support like credits of Cloud Storage or Cloud GPU to train my AI About my AI: NSAI-K2 is a custom-built AI assistant developed by Navtej Singh Saggar, powered by a self-designed Transformer language model, specialized AI agents, a custom training pipeline, and a…

KoFinClaim-Abstain: a Korean benchmark that scores LLMs by the cost of false accepts on financial numbers

Most LLM evals reward accuracy. In financial documents the expensive mistake is different: confirming a wrong number. I built KoFinClaim-Abstain — 500 Korean numeric claims from fictional filings. For each claim + 2–4 evidence sentences, a system must accept, reject, or hand off to a human. Scoring is by cost: a false accept costs 10, a review costs 1. Five open models answering alone passed 7–30% of wrong claims (EXAONE 3.5 7.8B: 36/500, Qwen3 8B: 149/500). Adding a rule-based checker that…

I Built GLITCH Studio — A Local AI Filmmaking Suite (Beta TesteProof — Seeking Independent Technical Feedback 18 114 14h Embedding a local LLM directly into an indie interrogation game 3 263 23h Can we prove this: the code submitted with this rs Wanted)

Meet GLITCH Studio — Local AI Filmmaking, Without the Credits Hi everyone! I’m an independent developer from New Zealand, and I’ve spent over a year building GLITCH Studio. What started as a personal creative project has grown into a Windows AI filmmaking suite, bringing powerful local AI tools together in one interface. GLITCH combines image generation, character creation, wardrobe, scenes, animation, talking characters and AI video workflows, using models including FLUX, WAN, MiniMax H3 and…

SaluNet: Can Learned Geometry Replace Conventional Feature Normalization?

I would like to highlight an interesting research contribution by Prof. Mourad Zaied, introducing SaluNet, a framework that explores an alternative perspective on normalization in modern deep learning architectures. The central idea is particularly interesting: instead of relying on batch-wise or activation statistics, can learnable transformations provide an effective alternative to conventional normalization layers? The proposed framework introduces three key components: SALU: A learnable…

[ENDORSEMENT HELP] For 1st time arxiv submission

Hey, all. I am a final year student, into AI research for the past 2 years. One of my paper got accepted into ACCV MM4SciReal 2026 workshop. But due to high registration fees and limited resources, i haven’t done registration. Now i want to submit the work into arxiv. As a first timer for arxiv submission, it would be very helpful if anyone among you guys can endorse me. This will take only some seconds. I have attached by endorsement code. Would be very helpful if you help in this. Regards, 1…

A question on the code used in AI

Why is it using higher level languages like Python, that are very slow to process and not Assembly or C for both size and optimisation. Using lower levels makes size smaller and usage for CPU, GPU, RAM becomes lower. For example Hello world in assembly or C takes 5seconds to proces while in python it takes around a min. I tryed to find information but the only thing I found was Higher levels are easier to read. 2 posts - 2 participants Read full topic

STDP + winner-take-all over 8 seeds: weight separation stuck near 1.9x (and a Brian2 warning we ignored)

We keep small spiking-network experiments that run on a laptop CPU (Brian2). Each learning experiment has a pass/fail bar set before it ran, and the nulls are kept: spiking-experiments. A warning first, because it cost us a wrong finding. Our exp5/exp6 kept two Synapses objects in a Python list. Brian2’s magic run() only collects objects it can see by name, so it skipped the noise->output synapses and said “never included in a network” on every run. We read past it and reported a “dead third…

Separating retrieval failures from generation failures in a RAG app

A team built a question-answering assistant over internal policy documents. Users kept reporting wrong answers, and the team’s first instinct was to try a bigger model. Before changing anything, they labeled about 200 failed conversations and sorted each into one of two groups: the right passage was never retrieved, or the right passage was retrieved but the model answered incorrectly. The split was roughly 70/30, with most failures coming from retrieval. Many policy documents had been chunked…