PIRL: From Open-Loop Exploration to Closed-Loop Reinforcement Learning [R]
TL;DR: Most RL post-training algorithms optimize the current batch and move on. But after an update, did the new policy actually become better? We introduce Policy…
Latest
NeurIPS 2026 AI-generated reviews [D]
I'm really confused about what the point of the prompt injection was (speaking as an author). Is it just a study? I would really prefer that…
How exactly does the NeurIPS meta reviewer response work? [D]
We can currently answer the meta-reviewer via AC confidential comment, but they just posted the following tweet: https://x.com/neuripsconf/status/2081991451236319328?s=46&t=HWfJoLgHxGH2W5l-o3mPJw Where it says: “Please post your answers to…
Are single GPU research still published in ML/DL and its applications nowadays? Which are the most notable recent ones? [D]
ML research is progressing at breakneck speed where frontier labs in both academia and industry have access to considerably large computes (GPUs). Where do small labs…
Pattern Recognition (Elsevier): "With Editor" status date changed, but status didn’t. Is this normal? [R]
Hi everyone, I have a manuscript under review at Pattern Recognition (Elsevier), and I'm a bit confused about the Editorial Manager status. My timeline is: Submitted:…
Editing Neurips Rebuttal [D]
It looks like the post rebuttal button will change to official comment July 27 AoE. Will I be able to edit my rebuttal once that happens?…
Evaluated 6 frontier LLMs (GPT-5.4, Claude Sonnet 4.6, Claude Opus 4.7, Gemini Pro/Flash, Grok 4.3) on political, gender, and racial bias across 8 benchmarks (~20,600 examples) [R]
I ran a solo evaluation project benchmarking six current frontier models: GPT-5.4, Claude Sonnet 4.6, Claude Opus 4.7, Gemini Pro, Gemini Flash, and Grok 4.3. I…
Training data needs a real go/no-go gate before training [D]
We have gates for code, infrastructure, deployment and model performance. But when it comes to the actual training artifact, the decision to proceed is often still…
Built & Trained a Transformer from Scratch in Pure PyTorch for English-to-Tamil Machine Translation [Math + Code Breakdown] [P]
Hi everyone! 👋 I built and trained the complete Transformer architecture from scratch using pure PyTorch (`torch.nn` primitives) based on the original "Attention Is All You…
Made a small model that extracts text from a white background [P]
Hello, I read a paper on a model named DONUT that extracts text from documents, which became my inspiration for this little project. Initially I wanted…