Skip to content
K = 0.731
All news · Today8 items

#Reasoning

All collected and published items, ordered by time. Picks are a selected subset.
9/27Sun
  1. 量子位40

    Fermi Universe launches FermiQLLM 1.0, a fully quantum-enhanced foundation model, with a RMB 1 billion valuation

    Tsinghua-affiliated startup Fermi Universe has launched FermiQLLM 1.0, which it calls the world's first fully quantum-enhanced foundation model. Adapted from the open-source Qwen base model, it applies quantum enhancement across data representation, model architecture, training, reinforcement, and evaluation. It can be trained and deployed on existing GPUs without fault-tolerant quantum computers.

9/26Sat
  1. Kevin Weil · X79

    Anthropic reports Claude solved a nine-loop calculation in planar N=4 super-Yang-Mills

    Anthropic reports that Claude solved a nine-loop scattering-amplitude problem in planar N=4 super-Yang-Mills, running largely unsupervised for days in Claude Science at a total cost of a few thousand dollars. Claude used methods developed by Lance Dixon and colleagues, and Dixon independently verified the result. Kevin Weil, who calculated Feynman diagrams in graduate school, calls the result mind-boggling and expresses excitement about where AI could take high-energy physics over the coming 12 months.

    Why it matters: The reported cost of a few thousand dollars offers a concrete reference point for assessing AI-assisted theoretical physics calculations within an academic compute budget.

9/24Thu
9/23Wed
9/22Tue
  1. Latent Space78

    Xiaomi releases MiMo-V2.6-Pro 1T-A42B, an open-source multimodal model trained for $3 million

    Xiaomi has released the MiMo-V2.6 family, including the multimodal MiMo-V2.6-Pro and MiMo-V2.6-Flash, and introduced MiMo-V2.6-Pro-UltraSpeed, which delivers up to 20× faster output.

    Why it matters: Xiaomi's first entry in the frontier open-model lineup, with open RL environments and training recipes, provides a concrete example of post-training cost strategies.

  3. Latent Space76

    TypeSafe AI founder Diogo Almeida on Jev and RLCD: building System 1 models for software, not chat

    In a Latent Space interview, TypeSafe AI founder Diogo Almeida introduces Jev as a System 1 model designed for code consumption and optimized for intelligence per dollar rather than chat experience. Its key innovation is unpublished RLCD (Reinforcement Learning for Calibrated Decisions), which uses calibrated probabilities instead of human-rating feedback.

    Why it matters: Almeida revisits the strategic split between RLHF and RLCD and explains why the model is optimized for reliability rather than benchmark scores.

9/19Sat
9/17Thu
9/16Wed
  1. NVIDIA Blog62

    NVIDIA Vera Rubin NVL72 makes its MLPerf Inference v6.1 debut with leading performance

    NVIDIA submitted preview results for Vera Rubin NVL72 for the first time in MLPerf Inference v6.1, reporting up to 3.7× the throughput of GB300 NVL72 on Qwen3-VL and up to 2.5× on DeepSeek-R1.

    Why it matters: MLPerf Inference v6.1 reports throughput, scaling efficiency, and software-optimization gains for Vera Rubin NVL72 versus GB300 NVL72, offering a basis for judging inference infrastructure value.

  3. Google Research36

    Google Research proposes Retrieve-for-Train: compile diffusion models with RL to bypass inference bottlenecks in complex AI search

    Google Research proposes Retrieve-for-Train, a framework that uses offline RL to discover reward-aligned query fan-outs and compile them into supervision, then distills them into a 53.9M-parameter diffusion retriever. At inference, it generates a complete result set in one non-autoregressive pass without chain-of-thought tokens. The method fine-tunes fan-out language models based on Gemma3-4B and Qwen3-4B, using set-level attribute rewards to address paraphrase collapse in zero-shot LLMs and autoregressive latency.

  4. Google DeepMind · X77

    Google DeepMind releases Gemini 3.8 Live and 3.8 Live Extended Thinking

    Google DeepMind has released Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, calling them its best conversational AI yet. The company says the models can converse and think while handling tasks in the background without interrupting the user's workflow.

    Why it matters: The new conversational models give readers an update on Google DeepMind's positioning for real-time dialogue and background task handling.

  5. NVIDIA Blog72

    NVIDIA shares efficiency progress for Vera Rubin and DSX at AI Infra Summit

    At the AI Infra Summit, NVIDIA announced several advances in Vera Rubin and its DSX platform, shifting AI infrastructure measurement from peak performance to useful agentic tokens per megawatt.

    Why it matters: The article reports several measured token-output figures per megawatt for Vera Rubin and DSX, showing how efficiency metrics for AI factories are changing.

9/12Sat
  2. Thinking Machines · X34

    Thinking Machines' John Schulman discusses where human judgment still matters as models self-improve

    Our own @johnschulman2 talks with Dwarkesh about where human judgment still matters as models improve and self-improve: teaching them to handle messy real-world tasks, applying taste to what works in the long run, and, above all, specifying what we actually want. Quoted post by @dwarkesh_sp (context, not the main author's work): New episode with @johnschulman2, @oneill_c and @BerenMillidge. I got together with some of the most insightful AI researchers I know who are at the openish companies, because I wanted to hear the details of what's actually happening at the frontier and what comes next. 0:00:00 – Steelmanning the case against RSI 0:18:39 – What's driving the Chinese labs' progress 0:28:06 – How will automated AI researchers be trained 0:33:51 – Will long-horizon RL elicit AGI? 0:45:24 – The sim-to-real gap 1:00:33 – How much progress is explained by data? 1:18:03 – Why is RL working so well? 1:24:54 – Move 37 and entropy collapse 1:28:31 – Rapid-fire timelines

9/10Thu
  1. DeepSeek · X68

    DeepSeek releases DeepSeek-V4.1-Flash

    DeepSeek has released DeepSeek-V4.1-Flash, describing it as smarter, faster, and more efficient. It is the smallest model in a new architecture family, with native visual understanding, and is designed for stronger capabilities, faster inference, and higher throughput, with the architecture able to scale to larger models.

    Why it matters: As the smallest model in DeepSeek's new architecture family, its native visual understanding and higher inference throughput offer a view of its lightweight multimodal positioning.

9/9Wed
9/8Tue
9/4Fri
  1. Sébastien Bubeck · X31

    Meet Astra's TikZ unicorn. I find it mind blowing that a single entity can do this, as well as get essentially 100% on Frontier Math Tier 4, ARC-AGI 3, & ExploitBench. And it's not behind closed doors, anyone can just go and talk to it. Impossible to overstate the implications.

    Meet Astra's TikZ unicorn. I find it mind blowing that a single entity can do this, as well as get essentially 100% on Frontier Math Tier 4, ARC-AGI 3, & ExploitBench. And it's not behind closed doors, anyone can just go and talk to it. Impossible to overstate the implications.

9/3Thu
8/25Tue
  1. Hugging Face Blog62

    IBM releases Granite 4.2 reasoning models in 3B, 8B, and 30B sizes

    IBM's Granite team has released Granite 4.2, its first family of dense decoder-only reasoning models, in 3B, 8B, and 30B sizes. They were post-trained from the Granite-4.1 base and are all licensed under Apache 2.0.

    Why it matters: The article discloses Granite 4.2's architecture, SFT data mix, and multi-stage RL curriculum in detail, offering a look at concrete post-training methods for reasoning models.

  2. Hugging Face Blog60

    Multiverse Computing proposes Quantization-Aware Healing: 4-bit compressed model beats its bfloat16 original on seven of nine benchmarks

    Multiverse Computing has published a paper proposing Quantization-Aware Healing (QAH). After compressing GPT-OSS 120B to 60B parameters and quantizing it to MXFP4, the method distills directly from the original model before compression rather than from a recovered bfloat16 checkpoint.

    Why it matters: The paper presents a repair recipe that distills from the original model after compression and quantization, with stability comparisons against QAT.

8/21Fri
8/17Mon
8/13Thu
  1. DeepSeek · X82

    DeepSeek releases DeepSeek-V4-Pro

    DeepSeek has released DeepSeek-V4-Pro, highlighting upgraded agent capabilities and notable production gains. V4-Pro and V4-Flash offer flexible reasoning levels: low for simple tasks, high for everyday agent workflows, and max for complex tasks. The model natively supports the OpenAI Responses API and is optimized for Codex with one-click configuration. It is now available in the app, on the web in Expert Mode, and through the API; its model name is unchanged.

    Why it matters: The official announcement describes V4-Pro's agent upgrades and three reasoning levels, helping readers understand its position in production workflows.

  2. Sébastien Bubeck · X58

    Sébastien Bubeck reacts to reported ChatGPT 5.6 solution of an open math problem

    Sébastien Bubeck calls the reported result amazing, noting that he spent a week trying to solve the problem back in 2012. In the quoted post, Steven Strogatz says a neurosurgery resident with no training in advanced math used ChatGPT 5.6 to solve a major open problem in numerical linear algebra. https://alextownsend.net/essays/SIAMNews_CrouzeixConjecture.pdf

  3. Microsoft Research36

    MindTopo exposes spatial-reasoning gaps in multimodal LLMs

    Microsoft Research has introduced MindTopo, a benchmark for topological reasoning in multimodal LLMs. It covers connectivity, separation, ordering, containment, and knot relations, across static reasoning and interactive planning. Current models are substantially better at static recognition than interactive planning, but both remain far below human performance; failures occur more often in planning than perception. Image and video generation helps only when topology is preserved within a single frame and remains unreliable across multi-step operations.

8/12Wed
  1. Google Research60

    Google Research proposes knowledge profiling: retrieval, not encoding, is the bottleneck for frontier LLM factuality

    Google Research has proposed a knowledge-profiling framework that distinguishes factual errors caused by encoding failures from those caused by retrieval failures, and released WikiProfile, a benchmark containing 2,150 Wikipedia facts.

    Why it matters: By separating encoding failures from retrieval failures, knowledge profiling offers a new way to diagnose the sources of factual errors in frontier models.

7/31Fri
  1. Mira Murati · X73

    Thinking Machines releases Inkling-Small with full weights at a quarter of Inkling’s size

    Mira Murati says Inkling-Small is comparable to Inkling at a quarter the size, with open weights and fine-tuning on Tinker today. She looks forward to seeing what people make with it. In the quoted announcement, @thinkymachines reports 276B total parameters and 12B active, makes the full weights available, and offers text, image, and audio chat on Tinker Playground. https://thinkingmachines.ai/news/inkling-small/

    Why it matters: The comparison with Inkling provides context for the smaller model’s claimed performance, while the parameter counts clarify the scale of the size reduction.

7/29Wed
  1. Berkeley AI Research62

    From CUDA to MLX: how K-Search brings kernel-optimization experience to Apple Silicon

    Berkeley AI Research and IBM Research have extended the evolutionary kernel-search framework K-Search to MLX and added a structured CUDA-to-MLX translation layer that automatically adapts existing CUDA-kernel knowledge to Apple Silicon.

    Why it matters: The article details the CUDA-to-MLX translation layer and reports two kernel benchmarks, showing how kernel knowledge can be transferred across platforms.

7/26Sun
  1. Berkeley AI Research36

    Berkeley AI Research proposes ABBEL, using belief-state supervision to make LLMs efficient at long-horizon interaction

    Berkeley AI Research has proposed ABBEL, a framework that separates an LLM's recursive summaries into natural-language “belief states” and scores their content with supervised signals to reduce performance loss from context compression in long interactions. On the collaborative-programming benchmark CollabBench, reconstruction-based belief scoring cut the performance gap with a full-context model by about 50%, reduced training from 100 steps to 50, and kept peak context length well below the full-context setup.

7/16Thu
7/15Wed
  1. Hugging Face Blog88

    Thinking Machines releases Inkling family of open multimodal models

    Thinking Machines Lab has released the open-source Inkling family on Hugging Face. Inkling is an MoE model with about 1T total and 41B active parameters, natively supporting image, audio, and text inputs with a 1M-token context. The release also includes Inkling-Small, with 276B total and 12B active parameters.

    Why it matters: The article gives Inkling's architecture, deployment setup, and benchmark comparisons, helping readers assess the practical requirements of an approximately 1T-parameter open multimodal model.

7/10Fri
7/2Thu
  1. Mistral AI71

    Mistral releases Leanstral 1.5 formal-verification model

    Mistral has released Leanstral 1.5, a formal-verification model licensed under Apache 2.0 with 119B total and 6B active parameters. It scores 100% on miniF2F, solves 587 of 672 PutnamBench problems, and sets state-of-the-art scores of 87% on FATE-H and 34% on FATE-X.

    Why it matters: The source reports Leanstral 1.5's formal-verification benchmark results and open-source availability, helping readers assess its usefulness in proof engineering.

6/27Sat
  1. Google Research60

    Google speeds up Gemini Nano on Pixel with frozen multi-token prediction

    Google Research has introduced an architecture that attaches a multi-token-prediction (MTP) head to the frozen Gemini Nano v3 model, speeding up on-device generation without changing the backbone weights.

    Why it matters: Google attaches an MTP head to frozen Gemini Nano v3 and reports concrete methods and gains under on-device memory and energy constraints.

6/25Thu
  1. Google Research37

    Google Research studies how reasoning unlocks LLMs' parametric knowledge

    A paper from Google Research at COLM 2026 finds that reasoning traces can unlock parametric knowledge in LLMs that they could not otherwise access, even for simple single-hop factual questions. Tests on Gemini-2.5 Flash and Pro and Qwen3-32B identify two mechanisms: extra tokens act as a “computation buffer,” and “fact priming” builds a retrieval bridge by generating related facts.