NeurIPS 2025, as a premier annual event in machine learning and computational neuroscience, tackles major topics like the future of AI, current research, and the most difficult challenges. While we’re not attending this year, we’re closely following the updates and today we pull together a quick, easy-to-digest roundup of a few standout papers so you can jump in without getting overwhelmed.
Here is a list of 15 papers from NeurIPS 2025, including 8 top research papers that received awards, along with 7 others that caught our attention:
1. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks → https://neurips.cc/virtual/2025/loc/san-diego/test-of-time/128328 Test of Time Award winner. Introduces the RPN, a small convnet that predicts objectness and boxes on shared features, enabling Faster R-CNN to share computation and run around 5 fps on a GPU
2. Artificial Hivemind: The Open-Ended Homogeneity of LMs (and Beyond) → https://neurips.cc/virtual/2025/loc/san-diego/poster/121421 Releases a huge open-ended prompt dataset, showing that LLMs often fall into an “artificial hivemind” – generate surprisingly similar answers – and measuring diversity collapse
3. Optimal Mistake Bounds for Transductive Online Learning → https://neurips.cc/virtual/2025/loc/san-diego/poster/119098 Settles a 30-year-old question by showing how much unlabeled data helps in online learning – it gives a precise quadratic advantage with tight matching bounds
4. Gated Attention for LLMs: Non-linearity, Sparsity, and Attention-Sink-Free → https://neurips.cc/virtual/2025/loc/san-diego/poster/120216 Demonstrates how gating actually affects attention: a simple sigmoid gate after Scaled Dot-Product Attention (SDPA) boosts performance, stability, and long-context behavior by adding useful nonlinearity and sparse modulation
9 Recent advances in Multi-Agent Systems (all open-source)
The idea to split tasks across multiple agents instead of relying on one universal agent is now seen as one of the most effective ways to build an AI stack. Concepts like “agent swarms” were highlighted at the AI Engineer Code Summit in NYC (Nov 20–21) as the winning architecture. And this trend is not only about coding and software. It applies across all AI domains.
So here is some recent research that helps keep multi-agent systems (MAS) better and up-to-date:
1. LatentMAS → Latent Collaboration in Multi-Agent Systems (2511.20639) AI agents share their hidden "thoughts" directly in latent space instead of talking through text. This makes collaboration and reasoning way faster and accurate (no extra training needed)
2. Puppeteer → Multi-Agent Collaboration via Evolving Orchestration (2505.19591) Uses a “puppeteer” LLM that dynamically decides which agents (“puppets”) to call and in what order. By learning this orchestration with reinforcement learning (RL), the system solves complex tasks more efficiently and with fewer compute costs
3. MADD → MADD: Multi-Agent Drug Discovery Orchestra (2511.08217) A MAS with 4 agents for drug discovery. It lets researchers describe a drug discovery task in plain language. Then MADD automatically builds and runs the full hit-identification pipeline, making AI-driven drug design a simple end-to-end workflow
4. Multi-Agent Tool-Integrated Policy Optimization (MATPO) → Multi-Agent Tool-Integrated Policy Optimization (2510.04678) Lets one LLM act as multiple agents (like a planner and a worker) by using different prompts and training them together with RL. So you get the benefits of a multi-agent system without needing multiple models
In AI, spatial intelligence is basically the model’s “sense of space” – its ability to understand where things are, how they relate, and how they move. It lets an AI models navigate a room, interpret a scene, or figure out how objects fit together, like giving it a built-in mental map. For example, world models can't live without spatial intelligence.
Here are 6 good reads to explore what spatial intelligence is and how it's evolving:
1. From Words to Worlds: Spatial Intelligence is AI’s Next Frontier by Fei-Fei Li → https://drfeifei.substack.com/p/from-words-to-worlds-spatial-intelligence Fei-Fei Li, the godmother of AI, is a key figure in spatial intelligence, since her work in computer vision, especially ImageNet, helped AI learn to recognize and understand objects in space. She's recently started a blog, and this post, in particular, argues that true intelligence requires grounding in space, understanding geometry, motion and consequences in the real world
2. Spatial Reasoning in Multimodal LLMs: A Survey of Tasks, Benchmarks and Methods → https://arxiv.org/abs/2511.15722 Breaks down how AI models handle spatial reasoning from a cognitive angle, maps all the existing tasks and benchmarks to that framework
3. What is Spatial Intelligence? → https://www.turingpost.com/p/cvhistory5 Our special article easily explains what spatial intelligence actually is, why it matters, and how researchers are trying to boost it so machines can better understand and navigate the physical world
4. From 2D to 3D Cognition: A Brief Survey of General World Models → https://arxiv.org/pdf/2506.20134 Shows how AI world models are evolving from simple 2D perception to full-on 3D understanding, explaining the tech behind it, what new 3D abilities these models gain, and where they’re used in the real world
Since Yann LeCun together with Randall Balestriero released a new paper on JEPA (Joint-Embedding Predictive Architecture), laying out its theory and introducing an efficient practical version called LeJEPA, we figured you might need even more JEPA. Here are 7 recent JEPA variants plus 5 iconic ones:
6. TS-JEPA (Time Series JEPA) → Joint Embeddings Go Temporal (2509.25449) Adapts JEPA to time-series by learning latent self-supervised representations and predicting future latents for robustness to noise and confounders
Precision is very important in AI as it shapes how accurate and efficient models are. It controls how finely numbers are represented, approximating real-world values with formats like fixed-point and floating-point. A recent BF16 → FP16 study renewed attention to precision impact. Here are the main precision types used in AI, from full precision for training to ultra-low precision for inference:
1. FP32 (Float32): Standard full-precision float used in most training: 1 sign bit, 8 exponent bits, 23 mantissa bits. Default for backward-compatible training and baseline numerical stability
2. FP16 (Float16) → https://arxiv.org/abs/2305.10947v6 Half-precision float. It balances accuracy and efficiency. 1 sign bit, 5 exponent bits, 10 mantissa bits. Common on NVIDIA Tensor Cores and mixed-precision setups. There’s now a new wave of using it in reinforcement learning: https://www.turingpost.com/p/fp16
Policy optimization (PO) algorithms are central to training AI models with preference-based feedback. In recent weeks, numerous new PO methods have emerged that build on or replace the popular PPO and GRPO, solving their issues. Here are 11 of them:
3. Asymmetric Importance Sampling Policy Optimization (ASPO) → ASPO: Asymmetric Importance Sampling Policy Optimization (2510.06062) Fixes imbalanced token weighting in LLM training. It flips the importance sampling ratios for positive tokens to correct over- and under-updates, and adds a soft dual-clipping step to keep gradients stable
4. In-Context Steered Policy Optimization (ICPO) → https://arxiv.org/abs/2510.26519 Uses a model’s own in-context learning ability to guide training with existing data. It combines Mixed-Policy GRPO with Implicit Expert Forcing to expand exploration and adds Expert Region Reject Sampling and Annealed Expert-Bonus Reward Shaping to ensure stability and balanced expert influence
5. Graph-Enhanced Policy Optimization (GEPO) → https://arxiv.org/abs/2510.26270 Builds a graph of an agent’s experiences to understand how different states connect, guide exploration and assign rewards more effectively
Coding is the field where AI is welcomed with open arms. Here’s a collection to help you take your AI-assisted coding workflows to the next level of convenience and efficiency:
1. Smol Developer → https://github.com/smol-ai/developer A lightweight AI “junior dev” that takes your product spec and automatically scaffolds or helps you build full codebases
2. Tabby → https://github.com/TabbyML/tabby A self-hosted AI coding assistant that runs locally as an alternative to GitHub Copilot. Easy to integrate, GPU-friendly, and doesn’t rely on the cloud
3. Beads (bd) Issue Tracker → https://github.com/steveyegge/beads Gives coding agents long-term memory, letting them organize, plan, and execute complex tasks reliably across sessions
4. MetaGPT → https://github.com/FoundationAgents/MetaGPT A multi-agent framework that imitates a software company team using LLMs. It assigns AI agents roles like PM, Architect, and Developer to produce user stories, designs, specs, and final code
5. Open Interpreter → https://github.com/openinterpreter/open-interpreter Gives you ChatGPT’s coding power with full local control – no limits, no sandbox – so you can automate, analyze, and create anything right from your desktop through a chat interface
6. OpenSpec → https://github.com/Fission-AI/OpenSpec A lightweight, spec-driven development tool that helps humans and AI agree on what to build before any code is written
7. PR-Agent → https://github.com/qodo-ai/pr-agent An AI code reviewer that automatically reviews, describes, and improves pull requests across GitHub, GitLab, and other platforms
8. BabyAGI → https://github.com/yoheinakajima/babyagi A self-building AI framework that gives agents the ability to write, manage, and refine their own functions, turning them from passive tools into active, self-building systems
If you want to understand the multifaceted AI landscape in 2025 and see where the field is heading – start with (or revisit) these legendary talks. They can help you capture what’s happening in AI from multiple angles:
1. Andrej Karpathy: Software Is Changing (Again) → https://www.youtube.com/watch?v=LCEmiRjPEtQ Unveils Software 3.0 – a paradigm where LLMs are the new computers, programmed with prompts instead of code. The key: developers must now master coding, training, and prompting as AI becomes the heart of software building
2. Richard Sutton, The OaK Architecture: A Vision of SuperIntelligence from Experience → https://www.youtube.com/watch?v=gEbbGyNkR2U Unveils the OaK (Options and Knowledge) architecture – a model-based RL framework for continual intelligence, where every component learns, meta-learns & builds hierarchical abstractions
3. GTC March 2025 Keynote with NVIDIA CEO Jensen Huang → https://www.youtube.com/watch?v=_waPvOwL9Z8 Dives into the accelerated computing and the importance of Physical AI. From the Blackwell GPU architecture & AI factories to breakthroughs in agentic AI & robotics, Jensen Huang explains how NVIDIA aims to power every layer of the AI ecosystem
4. Yann LeCun "Mathematical Obstacles on the Way to Human-Level AI" → https://www.youtube.com/watch?v=ETZfkkv6V7 Yann LeCun always argues we need a new path to machines that reason about the world – not LLMs or RL. So this lecture is about self-supervised systems with world models, planning, memory and energy-based learning
5. Andrew Ng: State of AI Agents → https://www.youtube.com/watch?v=4pYzYmSdSH4 Highlights one of the most pressing topics of 2025 – agents, explaining why most effective AI agents rely on simple, linear workflows built from modular “Lego-brick” tasks + what predicts AI startup success in the new agent era
Since Sora 2 is on fire these weeks, reminding us what high-quality video generation should look like, we decided you really need this list of video generation tools – great alternatives or complements to it.
1. Sora 2 → https://openai.com/sora/ It needs no introduction, but this OpenAI’s text-to-video model produces short, ultra-realistic clips across styles (cinematic, photorealistic, animated, etc.) with synced audio
2. Google Veo 3 (Gemini Video Generation) → https://aistudio.google.com/models/veo-3 Part of Gemini AI. Generates 8-second high-fidelity videos from text or images with native sound: background soundtracks and realistic voices with near-perfect lip sync
3. Runway (Gen-4 by Runway ML) → https://runwayml.com/ Text, image, or video-to-video generation with advanced editing like changing lighting, weather, camera angles or replacing objects. Popular in AI filmmaking
4. Pika Labs → https://pollo.ai/m/pika-ai Provides creative, often stylized short videos – from cinematic mini-scenes to cartoon-like animations. Ideal for social media clips and visual storytelling. Plus, you can add playful effects to manipulate objects in the generated videos
5. Luma’s Dream Machine → https://lumalabs.ai/dream-machine Powered by Luma AI’s latest Ray 3 model, it quickly visualizes story ideas, animated concept art, or abstract motion videos. It supports consistent custom characters and seamless looping
Reinforcement learning is having a moment - and not just this week. Some of its directions are already showing huge promise, while others are still early but exciting. Here’s a look at what’s happening right now in RL:
1. Reinforcement Pre-Training (RPT) → Reinforcement Pre-Training (2506.08007) Reframes next-token pretraining as RL with verifiable rewards, yielding scalable reasoning gains
2. Reinforcement Learning from Human Feedback (RLHF) → Deep reinforcement learning from human preferences (1706.03741) The top approach. It trains a model using human preference feedback, building a reward model and then optimizing the policy to generate outputs people prefer
The family of MCP (Model Context Protocol) servers keeps expanding to bridge agents, models, tools, web, data and apps. Here are 12 useful MCP servers that will help you create convenient agentic ecosystems:
1. Chrome DevTools MCP → https://github.com/ChromeDevTools/chrome-devtools-mcp Lets your coding agent (Gemini, Claude, Cursor, Copilot) control a live Chrome browser with full DevTools access for automation, debugging, and performance analysis
2. Windows-MCP → https://github.com/CursorTouch/Windows-MCP Provides interaction between agents and Windows, handling file navigation, app control, UI actions, QA testing
4. MetaMCP → https://github.com/metatool-ai/metamcp A proxy that aggregates multiple MCP servers into one, with middleware support. Works as a standard MCP server for any client
6. Playwright MCP → https://github.com/microsoft/playwright-mcp Lets LLMs interact with web pages via structured accessibility snapshots, no need for screenshots or visually-tuned models
8. Browserbase MCP Server → https://github.com/browserbase/mcp-server-browserbase Connects LLMs to external data and tools, adding cloud browser automation via Browserbase and Stagehand. It enables LLMs to browse, capture, extract, and act on web pages with precision
Low-Rank Adaptation (LoRA) is the go-to method for efficient model fine-tuning that adds small low-rank matrices instead of retraining full models. The field isn’t standing still – new LoRA variants push the limits of efficiency, generalization, and personalization. So we’re sharing 10 of the latest LoRA approaches you should know about:
4. aLoRA (Activated LoRA) → Activated LoRA: Fine-tuned LLMs for Intrinsics (2504.12397) Only applies LoRA after invocation, letting the model reuse the base model’s KV cache instead of recomputing the full turn’s KV cache. Efficient in multi-turn conversations
6 Recent & free sources to master Reinforcement Learning
Almost every week new research and resources on RL come out. Knowledge needs to be constantly refreshed and updated with the latest trends. So today, we’re sharing 6 free sources to help you stay on track with RL:
1. A Survey of Continual Reinforcement Learning → https://arxiv.org/abs/2506.21872 Covers continual RL (CRL): how agents can keep learning and adapt to new tasks without forgetting past ones. It analyses methods, benchmarks, evaluation metrics &challenges
3. Reinforcement Learning Specialization (Coursera, University of Alberta) → https://www.coursera.org/specializations/reinforcement-learning A 4-course series introducing foundational RL, implementing different algorithms, culminating in a capstone. It's a great structured path
5. A Survey of Reinforcement Learning for Software Engineering → https://arxiv.org/abs/2507.12483 Good if you're interested in RL-applied domains. Examines how RL is used in software engineering tasks: maintenance, development, evaluation. Covering 115 papers since DRL introduction, it summarizes trends, gaps & challenges
6. A Survey of Reinforcement Learning for LRMs → https://arxiv.org/abs/2509.08827 Tracks the way from LLMs to LRMs via RL. Covers reward design, policy optimization, use cases and future approaches like continual, memory, model-based RL and more
Models need feedback on what makes outputs “good” or “bad.” Policy optimization (PO) turns preferences and rewards into actual training signals. This field is evolving quickly, moving far beyond classics like PPO and GRPO. So here is our overview of 10 newest PO methods:
3. DCPO (Dynamic Clipping Policy Optimization) → DCPO: Dynamic Clipping Policy Optimization (2509.02333) Uses dynamic clipping, which adjusts probability limits per token for better token exploration, and smooth reward standardization to balance rewards over training steps and prevent wasted updates
4. ARPO (Agentic Reinforced Policy Optimization) → Agentic Reinforced Policy Optimization (2507.19849) Optimizes multi-turn LLM agents that use external tools. It uses an entropy-based adaptive rollout to explore post-tool use and an advantage attribution method to better assign credit across steps, leading to more efficient tool use with fewer resources
5. GRPO-RoC (Group Relative Policy Optimization with Resampling-on-Correct) → rStar2-Agent: Agentic Reasoning Technical Report (2508.20722) Oversamples rollouts, then resamples them to keep diverse mistakes and only the highest-quality correct answers. It reduces noises and ends up with stronger reasoning in a code environment
Everyone is buzzing around image generation this week, or more specifically, Google's Nano-Banana. So today we want to share a list of models that can be your great toolkit for image generation + editing + multi-turn refinement.
1. Gemini 2.5 Flash Image, or Nano-Banana → https://deepmind.google/models/gemini/image/ Google’s newest image model with conversational editing, character consistency, and multi-image fusion. Available in AI Studio and the Gemini API. Price: $2.50 per 1M tokens
2. FLUX (Black Forest Labs) → https://bfl.ai/ A family of models known for rich detail and, excellent prompt adherence, and fast iterative generation. Offered in several variants, from Pro to open-source, it's accessible via Hugging Face, Replicate, Azure AI Foundry, etc., and used as a base in many pipelines. Price: $0.025-0.08 per image
3. Midjourney v7 → https://www.midjourney.com/ Enhanced image fidelity, prompt comprehension, and anatomical coherence (hands, bodies, objects) + provides a smart lightbox editor. The Omni-reference tool improves character and object consistency in your images. It remains accessible via Discord with a supporting web interface. Price: $10-60/month
4. Stable Diffusion 3.5 (Stability AI) → https://stability.ai/stable-image Open-weights line with improved text rendering, photorealism, and prompt adherence compared to earlier versions. It introduces technical innovations through its MMDiT architecture. Price: $0.025-0.065 per image
5. OpenAI GPT-Image-1 →https://platform.openai.com/docs/guides/image-generation?image-generation-model=gpt-image-1 It's the same multimodal model that powers ChatGPT's image capabilities, offering high-fidelity image generation, precise edits, including inpainting, and accurate text rendering. Available via the Images API. Price: $40 per 1M tokens
Want to learn to build an AI Agent? I put together a cookbook for creating your own news research agent with OpenAI GPT-OSS:
- Searches headlines & specific sites - Pulls full articles when you need depth - Summarizes with clickable sources - Runs in a simple Gradio chat UI - No GPU, no local setup — just open-weight GPT-OSS models via Hugging Face
If you’ve been wanting to try agents but weren’t sure where to start, this is an end-to-end example you can fork, run, and adapt.
What can OpenAI’s new open models do with the news? I built a News Agent to find out.
It can answer questions about the news in real time, and every answer comes with original source links so you can dive deeper.
Ask it things like: - "What are the top news stories today?" - "What's the latest on artificial intelligence?" - Follow-up questions on specific stories
Runs with Hugging Face inference providers, letting you compare results from the OpenAI 20B and 120B models
So far, I’m quite impressed by the capabilities of even the smaller 20B model. Definitely not a perfect project, but curious to hear your thoughts!
OpenAI’s GPT-OSS has sparked ~400 new models on Hugging Face and racked up 5M downloads in less than a week, already outpacing DeepSeek R1’s first-week numbers.
For comparison: when R1 launched, I tracked 550 derivatives (across 8 base models) in a week, with ~3M downloads. GPT-OSS is ahead on adoption and engagement.
It’s also the most-liked release of any major LLM this summer. The 20B and 120B versions quickly shot past Kimi K2, GLM 4.5, and others in likes.
Most-downloaded GPT-OSS models include LM Studio and Unsloth AI versions: 1️⃣ openai/gpt-oss-20b - 2.0M 2️⃣ lmstudio-community/gpt-oss-20b-MLX-8bit - 750K 3️⃣ openai/gpt-oss-120b - 430K 4️⃣ unsloth/gpt-oss-20b-GGUF - 380K 5️⃣ lmstudio-community/gpt-oss-20b-GGUF - 330K
The 20B version is clearly finding its audience, showing the power of smaller, faster, more memory- and energy-efficient models. (These numbers don’t include calls to the models via inference providers, so the real usage is likely even bigger, especially for the 120B version)
Open-weight models let anyone build on top. Empower the builders, and innovation takes off. 🚀
Sharing some free, useful resources for you. In this collection, we’ve gathered the most recent books to give you up-to-date information on key fundamental topics. Hope this helps you master AI and machine learning:
1. Machine Learning Systems by Vijay Janapa Reddi → https://www.mlsysbook.ai/ Provides a framework for building effective ML solutions, covering data engineering, optimization, hardware-aware training, inference acceleration, architecture choice, and other key principles
2. Generative Diffusion Modeling: A Practical Handbook by Zihan Ding, Chi Jin → https://arxiv.org/abs/2412.17162 Offers a unified view of diffusion models: probabilistic, score-based, consistency, rectified flow, pre/post-training. It aligns notations with code to close the “paper-to-code” gap.
3. Geometric Deep Learning: Grids, Groups, Graphs, Geodesics, and Gauges → https://arxiv.org/abs/2104.13478 Explores unified geometric principles to analyze neural networks' architectures: CNNs, RNNs, GNNs, Transformers, and guide the design of the future ones
4. Mathematical Foundations of Geometric Deep Learning by Haitz Saez de Ocariz Borde and Michael Bronstein → https://arxiv.org/abs/2508.02723 Dives into the the key math concepts behind geometric Deep Learning: geometric and analytical structures, vector calculus, differential geometry, etc.
5. Interpretable Machine Learning by Christoph Molnar → https://github.com/christophM/interpretable-ml-book Practical guide to simple, transparent models (e.g., decision trees) and model-agnostic methods like LIME, Shapley values, permutation importance, and accumulated local effects.
6. Understanding Deep Learning by Simon J.D. Prince → https://udlbook.github.io/udlbook/ Explores core deep learning concenpts: models, training, evaluation, RL, architectures for images, text, and graphs, addressing open theoretical questions
