---
title: "AlphaSignal Reasoning: Math Reasoning News"
url: "https://alphasignal.ai/topic/reasoning/math-reasoning"
description: "Latest AI and machine learning news about Math Reasoning (Reasoning) on AlphaSignal."
---

# AlphaSignal Reasoning: Math Reasoning News

Latest AI and machine learning news about Math Reasoning (Reasoning) on AlphaSignal.

## Articles

- [OpenAI Drops 722 Math Papers From a Model Smarter Than GPT-6](https://alphasignal.ai/news/openai-drops-722-math-papers-from-a-model-smarter-than-gpt-6.md) (2026-10-06)
- [Stanford's Priced Guidance Framework Measures How Hard AI Finds Research Ideas](https://alphasignal.ai/news/stanford-s-priced-guidance-framework-measures-how-hard-ai-finds-research-ideas.md) (2026-10-04)
- [Meta's Muse Spark Helped Mathematicians Solve Five Open Research Problems](https://alphasignal.ai/news/meta-s-muse-spark-helped-mathematicians-solve-five-open-research-problems.md) (2026-10-02)
- [NVIDIA's Sigma Pushes Continuous Text Diffusion to 8 Billion Parameters](https://alphasignal.ai/news/nvidia-s-sigma-pushes-continuous-text-diffusion-to-8-billion-parameters.md) (2026-10-02)
- [Harvard's Projection Sampling Makes Standard Fine-Tuning Match Reinforcement Learning](https://alphasignal.ai/news/harvard-s-projection-sampling-makes-standard-fine-tuning-match-reinforcement.md) (2026-10-01)
- [Tencent's ExplorationBench Forces AI to Discover Hidden Rules Through Active Experimentation](https://alphasignal.ai/news/tencent-s-explorationbench-forces-ai-to-discover-hidden-rules-through-active.md) (2026-09-30)
- [AC2 Beats GRPO on Math Proofs Using 2.5x Fewer Decoding FLOPs](https://alphasignal.ai/news/ac2-beats-grpo-on-math-proofs-using-2-5x-fewer-decoding-flops.md) (2026-09-30)
- [UC Berkeley's EasyPPO Stops AI Training Collapses With Three Critic Fixes](https://alphasignal.ai/news/uc-berkeley-s-easyppo-stops-ai-training-collapses-with-three-critic-fixes.md) (2026-09-29)
- [Vals AI's Ten Claude Agents Formally Prove a Century-Old Math Problem](https://alphasignal.ai/news/vals-ai-s-ten-claude-agents-formally-prove-a-century-old-math-problem.md) (2026-09-29)
- [Vals AI's MysteryMechanism Benchmark Shows GPT-6 Astra Tops Science Test at 53%](https://alphasignal.ai/news/vals-ai-s-mysterymechanism-benchmark-shows-gpt-6-astra-tops-science-test-at-53.md) (2026-09-28)
- [Frontier Learning Keeps AI Training Sharp by Chasing Its Own Limits](https://alphasignal.ai/news/frontier-learning-keeps-ai-training-sharp-by-chasing-its-own-limits.md) (2026-09-28)
- [GPT, Gemini, and Claude All Pick the Same Right Answers](https://alphasignal.ai/news/gpt-gemini-and-claude-all-pick-the-same-right-answers.md) (2026-09-28)
- [Anthropic's Claude Breaks Physics Record With a Nine-Loop Particle Calculation](https://alphasignal.ai/news/anthropic-s-claude-breaks-physics-record-with-a-nine-loop-particle-calculation.md) (2026-09-25)
- [Recursive Self-Distillation Doubles Qwen3-8B Math Accuracy to 65.97%](https://alphasignal.ai/news/recursive-self-distillation-doubles-qwen3-8b-math-accuracy-to-65-97.md) (2026-09-25)
- [UCLA's PTTS Gives AI Parallel Reasoning Branches Separate Plans, Gaining 13.4 Points](https://alphasignal.ai/news/ucla-s-ptts-gives-ai-parallel-reasoning-branches-separate-plans-gaining-13-4.md) (2026-09-23)
- [Vals AI's Claude Opus 5.5 Agents Proved a Faster Shortest-Path Algorithm](https://alphasignal.ai/news/vals-ai-s-claude-opus-5-5-agents-proved-a-faster-shortest-path-algorithm.md) (2026-09-22)
- [Hill Sampling Beats AlphaEvolve on Math With 8 GPUs and Frozen Weights](https://alphasignal.ai/news/hill-sampling-beats-alphaevolve-on-math-with-8-gpus-and-frozen-weights.md) (2026-09-22)
- [Kyutai's Voice of Reason Thinks While It Speaks, Hitting 77% Math Accuracy](https://alphasignal.ai/news/kyutai-s-voice-of-reason-thinks-while-it-speaks-hitting-77-math-accuracy.md) (2026-09-21)
- [GPT-6 Astra Cracks a Decade-Old Voting Theory Problem Nobody Could Solve](https://alphasignal.ai/news/gpt-6-astra-cracks-a-decade-old-voting-theory-problem-nobody-could-solve.md) (2026-09-18)
- [Vals AI's MysteryMechanism Benchmark Tests If AI Can Rediscover 222 Scientific Laws](https://alphasignal.ai/news/vals-ai-s-mysterymechanism-benchmark-tests-if-ai-can-rediscover-222-scientific.md) (2026-09-16)
- [OpenAI's GPT-6 Astra Cracks Epoch's Hardest Math Benchmark in 14 Months](https://alphasignal.ai/news/openai-s-gpt-6-astra-cracks-epoch-s-hardest-math-benchmark-in-14-months.md) (2026-09-10)
- [Looped Flows Beats Rival Reasoners on ARC-AGI With Just 7M Parameters](https://alphasignal.ai/news/looped-flows-beats-rival-reasoners-on-arc-agi-with-just-7m-parameters.md) (2026-09-10)
- [Together AI Says Open-Source Kimi K3 Beats Claude Fable 5.1 on Legal Tasks](https://alphasignal.ai/news/together-ai-says-open-source-kimi-k3-beats-claude-fable-5-1-on-legal-tasks.md) (2026-09-10)
- [UT Austin Finds Fractal Chaos Explains Why AI Reasoning Costs Spike 10x](https://alphasignal.ai/news/ut-austin-finds-fractal-chaos-explains-why-ai-reasoning-costs-spike-10x.md) (2026-09-08)
- [OpenAI's 10,000-Agent Swarm Solves a 100-Year-Old Math Mystery](https://alphasignal.ai/news/openai-s-10-000-agent-swarm-solves-a-100-year-old-math-mystery.md) (2026-09-08)
- [Anthropic's Claude Formally Proved Fermat's Last Theorem in 11 Days](https://alphasignal.ai/news/anthropic-s-claude-formally-proved-fermat-s-last-theorem-in-11-days.md) (2026-09-04)
- [ReViSQL-K2.6 Beats Human Accuracy on SQL Benchmark Without Agentic Scaffolding](https://alphasignal.ai/news/revisql-k2-6-beats-human-accuracy-on-sql-benchmark-without-agentic-scaffolding.md) (2026-08-27)
- [CMU's Jeremy Avigad Says AI Math Is Bigger Than Theorem Provers](https://alphasignal.ai/news/cmu-s-jeremy-avigad-says-ai-math-is-bigger-than-theorem-provers.md) (2026-08-24)
- [DeepSeek V4 Pro Hits 90% on ARC-AGI-1 but Stalls on Harder Puzzles](https://alphasignal.ai/news/deepseek-v4-pro-hits-90-on-arc-agi-1-but-stalls-on-harder-puzzles.md) (2026-08-21)
- [Tencent's Hyra Agent Cracks Four Unsolved Math Problems at Once](https://alphasignal.ai/news/tencent-s-hyra-agent-cracks-four-unsolved-math-problems-at-once.md) (2026-08-19)
- [GPT-5 Proves Gradient Descent Hits a Hard Speed Limit at 1.9319](https://alphasignal.ai/news/gpt-5-proves-gradient-descent-hits-a-hard-speed-limit-at-1-9319.md) (2026-08-11)
- [Anthropic's Claude Pushes a 160-Year-Old Math Boundary From 41.6% to 67.2%](https://alphasignal.ai/news/anthropic-s-claude-pushes-a-160-year-old-math-boundary-from-41-6-to-67-2.md) (2026-08-10)
- [Epoch Hides a Game's Identity to Stop AI Labs From Cheating Benchmarks](https://alphasignal.ai/news/epoch-hides-a-game-s-identity-to-stop-ai-labs-from-cheating-benchmarks.md) (2026-08-06)
- [Meta AI Sweeps Five STEM Olympiads With Perfect Scores and Zero Tools](https://alphasignal.ai/news/meta-ai-sweeps-five-stem-olympiads-with-perfect-scores-and-zero-tools.md) (2026-08-06)
- [OpenAI's Astra Cracks 10 Math Problems That Stumped Humans for Decades](https://alphasignal.ai/news/openai-s-astra-cracks-10-math-problems-that-stumped-humans-for-decades.md) (2026-08-03)
- [Epoch Expands FrontierMath to 50 Unsolved Problems AI Has Already Cracked Three](https://alphasignal.ai/news/epoch-expands-frontiermath-to-50-unsolved-problems-ai-has-already-cracked-three.md) (2026-07-31)
- [Singapore's Agnes AI Builds a Reasoning Model That Uses 72% Fewer Tokens](https://alphasignal.ai/news/singapore-s-agnes-ai-builds-a-reasoning-model-that-uses-72-fewer-tokens.md) (2026-07-30)
- [Epoch's FrontierMath Benchmark Cracks a 40-Year-Old Math Problem Using Fable 5](https://alphasignal.ai/news/epoch-s-frontiermath-benchmark-cracks-a-40-year-old-math-problem-using-fable-5.md) (2026-07-28)
- [METR's Expenditure Horizon Reveals When AI Agents Beat Human Researchers](https://alphasignal.ai/news/metr-s-expenditure-horizon-reveals-when-ai-agents-beat-human-researchers.md) (2026-07-21)
- [Sakana AI's UnMaskFork Gives Diffusion Language Models Test-Time Scaling Power](https://alphasignal.ai/news/sakana-ai-s-unmaskfork-gives-diffusion-language-models-test-time-scaling-power.md) (2026-07-21)
- [DeepSeek AI's DeepSeekMath-7B Beats 540B Models and Birthed GRPO](https://alphasignal.ai/news/deepseek-ai-s-deepseekmath-7b-beats-540b-models-and-birthed-grpo.md) (2026-07-04)
- [DeepSeek-R1 Crashed Nvidia's Stock and Rewrote AI's Cost Rules](https://alphasignal.ai/news/deepseek-r1-crashed-nvidia-s-stock-and-rewrote-ai-s-cost-rules.md) (2026-07-04)
- [DeepSeek Releases DeepSeek-Prover-V2-7B to Write Verified Math Proofs on a Single GPU](https://alphasignal.ai/news/deepseek-releases-deepseek-prover-v2-7b-to-write-verified-math-proofs-on-a.md) (2026-07-04)
- [DeepSeek's R1-0528-Qwen3-8B Beats o3-mini on Math Running on a Gaming GPU](https://alphasignal.ai/news/deepseek-s-r1-0528-qwen3-8b-beats-o3-mini-on-math-running-on-a-gaming-gpu.md) (2026-07-04)
- [Epoch's EBR-Bench Shows GPT-5.5 and Claude Can't Learn From Practice](https://alphasignal.ai/news/epoch-s-ebr-bench-shows-gpt-5-5-and-claude-can-t-learn-from-practice.md) (2026-07-02)
- [Tufa Labs Wins $25K Beating Frontier AI on the World's Hardest Benchmark](https://alphasignal.ai/news/tufa-labs-wins-25k-beating-frontier-ai-on-the-world-s-hardest-benchmark.md) (2026-07-02)
- [OpenAI's o3 Cracked 18 Unsolvable Rare Disease Cases in Children](https://alphasignal.ai/news/openai-s-o3-cracked-18-unsolvable-rare-disease-cases-in-children.md) (2026-06-18)
- [Anthropic's Claude Fable 5 Hits 88% on Math's Hardest AI Benchmark](https://alphasignal.ai/news/anthropic-s-claude-fable-5-hits-88-on-math-s-hardest-ai-benchmark.md) (2026-06-12)
- [Epoch Caught 42% of FrontierMath Problems Broken, Sending GPT-5.5 Scores Soaring](https://alphasignal.ai/news/epoch-caught-42-of-frontiermath-problems-broken-sending-gpt-5-5-scores-soaring.md) (2026-06-12)
- [OpenAI's Reasoning Model Disproves an 80-Year-Old Math Conjecture Autonomously](https://alphasignal.ai/news/openai-s-reasoning-model-disproves-an-80-year-old-math-conjecture-autonomously.md) (2026-06-04)