Skip to content
K = 0.731
All news · Today8 items

#Reasoning

All collected and published items, ordered by time. Picks are a selected subset.
10/7Wed
  1. The Decoder80

    OpenAI posts 372 AI-generated mathematical proofs on GitHub

    OpenAI posted 372 mathematical results generated by an internal frontier model in a GitHub repository. It says each result solves or substantially advances an open problem, including improvements to major computational algorithms and progress related to the Riemann hypothesis.

    Why it matters: OpenAI posted 372 AI-generated mathematical results directly on GitHub rather than in a journal, highlighting tension between academic review processes and the speed of AI output.

  2. 量子位22

    Do any AIs still speak like people? Claude and GPT chain-of-thought reportedly use 'caveman language' to save tokens

    A QbitAI article complains that models such as Claude and GPT have become difficult to understand in coding contexts, using phrases such as 'ledger 15/15 all green' and 'one leg failed'. Online commenters suggest the main cause is chain-of-thought being compressed into 'caveman language' to save tokens, using 70% fewer tokens than plain language. The article says Opus 5.5 has improved its language ability, while Gemini is praised as one of the few models that still speaks naturally. CaveMan Skill, a GitHub project that teaches AI to speak in jargon, has 104K stars.

  3. Latent Space88

    OpenAI's internal model releases 722 math papers, claiming to solve 90 of 500 open problems

    OpenAI published mathematical results from an unreleased internal frontier model through a blog, GitHub repository, and posts. The release includes 722 manuscripts grouped into 372 related result families, drawn from an evaluation of about 4,000 research questions. Each result used an average of about three hours of ChatGPT Pro reasoning compute.

    Why it matters: OpenAI is producing mathematical results in bulk with an internal model, giving readers a view of the scale, compute cost, and outside criticism surrounding the work.

  4. 量子位85

    OpenAI releases 722 math manuscripts covering the Riemann, Hodge, and BSD conjectures

    OpenAI released 722 mathematical manuscripts grouped into 372 result clusters. They all came from an unreleased internal model, and the papers, source code, and some Lean proofs have been uploaded to GitHub. Topics include the quasi-Riemann hypothesis, Hodge conjecture, BSD conjecture, and the irrationality measure of π, spanning 17 areas. OpenAI says the model attempted about 4,000 research questions, with compute for each result averaging about three hours of ChatGPT Pro reasoning.

    Why it matters: OpenAI released 722 mathematical manuscripts at once across several areas related to Millennium Prize problems, while surfacing concerns in the mathematics community about verification and reading burden.

  5. Greg Brockman · X70

    OpenAI releases a batch of mathematical results from an internal frontier model

    OpenAI released new mathematical results produced by an internal frontier model, with related code and results at github.com/openai/math. OpenAI said it consulted an independent advisory group on mathematics and AI at the Institute for Advanced Study in Princeton about how to publish the work and considered its recommendations. Greg Brockman reposted the announcement, saying it points toward accelerating scientific discovery and improving quality of life for everyone.

    Why it matters: OpenAI made a batch of mathematical results from an internal frontier model public and said it followed advice from an external advisory group on the release.

  6. Kevin Weil · X50

    Kevin Weil praises OpenAI’s mathematics release and anticipates progress in physical sciences

    Kevin Weil celebrates OpenAI’s mathematics release today, saying much work remains to get models of the same caliber in the physical sciences. He urges imagining the possibilities and expresses confidence that models will get there. In the quoted post, @willdepue says he asked GPT 6 Pro and Fable 5.1 to rank discoveries from the last three years by human or AI origin, reporting that 81% were released today.

  8. OpenAI · X62

    OpenAI releases mathematical results from an internal frontier model

    OpenAI released a batch of mathematical results produced by an internal frontier model, with related code and materials hosted at github.com/openai/math. OpenAI said it consulted an independent advisory group on mathematics and AI at the Institute for Advanced Study in Princeton and considered its recommendations and public guidance when deciding how to publish the work.

10/6Tue
  1. Arthur Mensch · X72

    Mistral announces Mistral Large 4 with 1T parameters and native multimodality

    Mistral AI announced Mistral Large 4, aka Le Chonk, with 1T parameters, 49B active, and native multimodality, available via API today with an open weights release planned for the end of October. It claims the model leads open weights models from the US or Europe on aggregated benchmarks and surpasses closed frontier models on visual grounding. Arthur Mensch adds that it took groundwork, was trained and served on their own compute, and that RL shows no sign of saturation.

    Why it matters: The announced API availability and end of October open weights release give teams distinct timelines for accessing the model through hosted infrastructure or its weights.

  3. TechCrunch · AI72

    Reflection releases open-weight Beam model to compete with Chinese open models

    Reflection AI officially released its first frontier open-weight model, Beam, saying it performs comparably to leading Chinese open models on advanced reasoning benchmarks with 3–4 times less inference compute. Beam is a 501-billion-parameter, 23-billion-active-parameter text model, pretrained on 23.8 trillion tokens with a 1-million-token context window. The company says it matches Z.ai's GLM-5.2 and exceeds current leading Western open models.

  4. Simon Willison22

    Benchmark tests whether Qwen3.8 27B can add using English number words

    A benchmark tested the local Qwen3.8-27B-Q4_K_M.gguf on 5,070 cases without reasoning, asking it to add positive integers using only English words and return exact results. It achieved 23.57% numerical accuracy and 96.17% format compliance. Accuracy fell as the numbers grew, from 97.04% on one- to three-digit numbers to 6.44% on ten- to thirteen-digit numbers. In 169 paired cases with medium reasoning enabled, the model answered 167 correctly.

10/5Mon
  1. The Verge · AI22

    Our brains were not built for AI

    Neuroscientist Paul Cisek argues that the brain is not an information processor but a feedback-control system. A baseball outfielder catching a fly ball, for example, acts on the simple heuristic of 'keeping the ball in the same place in their field of vision' rather than calculating its trajectory. The article says AI is like 'cognitive junk food': immediately appealing but damaging to cognitive systems humans evolved over thousands of years.

10/3Sat
  2. AI at Meta · X66

    Meta publishes six papers on mathematicians collaborating with Muse Spark

    Meta said that after AI models achieved gold-medal-level performance in five math, physics, and chemistry competitions, it had mathematicians work with Muse Spark 1.1 and 1.2 in Thinking Mode through the meta.ai chat interface to solve open problems with no existing solution path and no custom research scaffolding.

    Why it matters: Meta made six papers from mathematicians' collaboration with Muse Spark public and described the division of work between people and AI and the review process.

10/2Fri
  1. MIT Technology Review · AI38

    Don't be fooled by LLMs: large language models do not truly reason

    A former Google DeepMind researcher says an LLM's chain of thought is simply an extended iteration of the same next-token prediction process, not an independent reasoning mechanism like AlphaGo's. He identifies three major limitations: no persistent, inspectable cognitive state; knowledge and reasoning tangled together in the weights; and chains of thought that are often fabricated after the fact. He argues for borrowing AlphaGo's game-tree architecture to give machine reasoning an explicit cognitive state.

  4. AWS Machine Learning Blog32

    How Amazon Payments uses contextual multi-armed bandits on Amazon SageMaker AI to improve acquisition funnel conversion

    Amazon Payments used a multi-objective contextual multi-armed bandit (LinUCB) on Amazon SageMaker AI to personalize content in its product-acquisition funnel. In a seven-week online A/B test, one audience segment achieved a high-single-digit relative increase in final funnel conversion, while another showed no improvement. The team ran one LinUCB model for each stage—application start, submission, and approval—and linearly combined their UCB scores to optimize the full funnel. It open-sourced a repository that can be tried with synthetic data.

  6. 量子位82

    Google launches Gemini 4 Argon, tops multiple leaderboards at about half Astra's cost

    Google released its next-generation flagship model, Gemini 4 Argon. It scored 77.9% on DeepSWE v1.1, ahead of Claude Opus 5.5 at 74.2% and GPT-6 Astra at 74.1%, and ranked first on Text Arena with 1,525 points.

    Why it matters: Google's new flagship tops several leaderboards and publishes cost and maximum output-token figures, helping readers assess its position against GPT and Opus.

  7. 量子位62

    Kaiming He's team proposes NAT-ARC: ImageNet pretraining for a purely visual ARC solver

    Kaiming He's team proposed NAT-ARC, a purely visual ARC-solving system that does not rely on LLMs. It pretrains an encoder with MAE on ImageNet and transfers it to abstract grid reasoning. The best single model has 0.6B parameters and reaches 63.4% pass@2 on ARC-1; an ensemble reaches 70.2%, with about one-quarter the parameters of the specialized LLM system The ARChitects, which uses an 8B model and reaches 71.6%.

10/1Thu
  2. NVIDIA Blog22

    How NVIDIA AI factories maximize ROI through high output, long life, and broad applicability

    NVIDIA says AI factory return on investment depends on three factors: output capacity, useful life, and demand. Its platform uses full-stack co-design to maximize throughput per megawatt. SemiAnalysis AgentX data show Vera Rubin NVL72 delivers more than 30 times the throughput per megawatt of GB300 NVL72 and, on DeepSeek V4 Pro, can reduce cost per million tokens by up to 45 times.

  3. The Decoder80

    OpenAI says it blocked model-reasoning theft, but the same method still works on Azure

    OpenAI said it blocked a distillation operation intended to steal model reasoning. Activity began at low volume on July 1 and surged on July 24–25 to 16,000 requests from more than 4,000 users across more than 15,000 linked accounts. OpenAI says it shut them all down by July 28, noting that these were attempted extractions and may not have succeeded.

    Why it matters: OpenAI disclosed a model-reasoning distillation theft incident that also raised questions about inconsistent cloud-platform protections, showing a real path for reasoning leakage.

  4. Latent Space82

    Google DeepMind releases Gemini 4 Argon with up to 1M output tokens

    Google DeepMind released Gemini 4 Argon for coding, enterprise knowledge work, and cyber defense. It is initially available to government users and trusted cybersecurity professionals in the Fairwind program, before expanding to developers, businesses, and consumers.

    Why it matters: This summary combines Gemini 4 Argon's benchmark results, pricing, and differences in third-party evaluations, enabling comparison with Astra and Opus 5.5.

  5. The Decoder78

    Google releases Gemini 4 Argon, matching OpenAI and Anthropic but without a clear lead

    Google released frontier model Gemini 4 Argon, matching OpenAI and Anthropic on several benchmarks without a clear lead. It is initially available to 'trusted cyber defenders' in the Fairwind program and Google's internal teams before expanding to developers, businesses, and consumers. Promotional API pricing is $2 per million input tokens and $10 per million output tokens.

    Why it matters: This summary compares third-party and human-preference leaderboards, scores, token use, and pricing for Gemini 4 Argon against OpenAI and Anthropic.

  6. Google DeepMind88

    Google DeepMind releases frontier model Gemini 4 Argon

    Google DeepMind released frontier model Gemini 4 Argon, initially making it available through the Fairwind Program to trusted cybersecurity defenders before gradually expanding access. Pricing starts at $2 per million input tokens and $10 per million output tokens.

    Why it matters: Gemini 4 Argon raises the maximum output to 1M tokens and reports multiple benchmark results and internal deployment data, offering a view of its limits on long-running workflows.

  8. Google AI · X62

    Google announces Gemini 4 Argon frontier model

    Google announced Gemini 4 Argon, a new frontier model built to sustain deep reasoning across complex, long-horizon workflows, with frontier performance in software engineering, enterprise knowledge work such as legal and finance, and cybersecurity defense. Google is expanding the model's output token limit to 1M tokens. Argon is rolling out to trusted cyber defenders in the Fairwind Program, with broader availability planned as soon as possible.

9/30Wed
  1. 腾讯混元 · X45

    Tencent Hy, Fudan, and Tsinghua Release ExplorationBench for AI Exploration

    Researchers from Tencent Hy, Fudan University, and Tsinghua University released ExplorationBench, a benchmark measuring how AI systems explore by discovering unfamiliar rules. It uses two verifiable "Alien Worlds": AlienCode (31 hidden rule changes, 70 tasks) and AlienLogic (24 patched inference rules, 70 theorems), graded by an interpreter or proof checker with no LLM judge. Across 10 frontier AI systems, feedback beat thinking alone, and rankings barely transferred between the two worlds.

  3. 雷锋网40

    Why do Kimi and DeepSeek appear in the same model-architecture diagram?

    Of the 45 layers in GLM-5.3-Flash, 34 use Kimi's KDA linear-attention approach and 11 use DeepSeek's KPool-DSA sparse attention. Linear and Sparse attention appear together in one model for the first time. The combination reflects a shift away from finding a single best mechanism toward assigning different roles: Linear maintains long histories at low cost, while Sparse retrieves specific information as needed.

  5. 雷锋网28

    From 3D vision to world models: why spatial intelligence is a common thread in AI's move into the physical world | ECCV 2026 recap

    ECCV 2026 received 10,473 submissions and accepted 2,834. The 3D topic fell from a leading area to third place, but questions around 3D representation, geometry, and physical consistency are spreading into video generation, embodied AI, and world models. Of 93 workshops, about 10 were directly related to 3D, 11 to spatial intelligence, nine to physical AI, and five to world models. Best-paper and honorable-mention awards also focused on 3D surface representation, object reconstruction, and geometry perception.

9/29Tue