#reasoning

27 Aug 2026 · Paper summary
Paper summary: VGI-Bench: Probing Visual Intelligence in Video Generation Models
VGI-Bench tests visual reasoning in video generation models with 27 tasks and 810 instances that require a valid process, not just a plausible final frame; the best model, Seedance 2.0, scores 51.0%.
12 Aug 2026 · Study
Agentic AI · ReAct
ReAct interleaves reasoning ("Thought") with tool use ("Action") and results ("Observation"); the prompt format, a parser for model output, and a full trace with a scripted model.
11 Aug 2026 · Paper summary
Paper summary: BDH-CQ: In-Context Learning with Recurrent Latent Reasoning
BDH-CQ combines in-context learning with recurrent latent reasoning in the Dragon Hatchling architecture; a 150M model reaches 29.5% pass@2 on ARC-AGI-1 at about $0.0007 per task, a new cost–accuracy frontier.
29 Apr 2026 · Paper summary
Paper summary: Recursive Multi-Agent Systems
RecursiveMAS treats a whole multi-agent system as one recursive computation in latent space, linking agents with a light RecursiveLink module and training them jointly; it reports +8.3% accuracy and 1.2–2.4× faster inference.
24 Feb 2026 · Paper summary
Paper summary: A Very Big Video Reasoning Suite
VBVR is a video-reasoning dataset of 200 curated tasks and over a million clips, about 1,000× larger than earlier ones, with a rule-based benchmark used for an early study of how video reasoning scales.
19 Feb 2026 · AI updates
Gemini 3.1 Pro: upgraded core reasoning
Google released Gemini 3.1 Pro in preview, the core model behind the previous week's Gemini 3 Deep Think update, rolling out across the Gemini API, Antigravity, Vertex AI, the Gemini app and NotebookLM.
22 Jan 2026 · Paper summary
Paper summary: A Survey of Agentic Reasoning for Large Language Models: Towards Recursively Self-Improving and Collective Agents
A survey organising agentic reasoning into three layers (foundational single-agent skills, self-evolving agents, collective multi-agent reasoning), each split into in-context and post-training approaches.
7 Nov 2025 · Paper summary
Paper summary: Thinking with Video: Video Generation as a Promising Multimodal Reasoning Paradigm
"Thinking with Video" uses a video generation model (Sora-2) as a reasoner, drawing its solution as video frames; on a new benchmark it rivals strong VLMs on visual puzzles and reaches 92% on MATH.
30 Oct 2025 · Paper summary
Paper summary: Scaling Latent Reasoning via Looped Language Models
Ouro builds latent reasoning into pre-training with looped language models (iterating in latent space with a learned depth), and its 1.4B and 2.6B models are reported to match up to 12B models.
20 Oct 2025 · Paper summary
Paper summary: A Theoretical Study on Bridging Internal Probability and Self-Consistency for LLM Reasoning
A theoretical framework for sampling-based test-time scaling shows self-consistency has high estimation error and perplexity has high modelling error; the hybrid RPC method matches self-consistency with half the samples.
8 Oct 2025 · Paper summary
Paper summary: Less is More: Recursive Reasoning with Tiny Networks
Tiny Recursive Model, a single 2-layer network with 7M parameters that repeatedly refines its answer, reports 45% on ARC-AGI-1 and 8% on ARC-AGI-2, above many large LLMs on these puzzles.
19 Sept 2025 · Paper summary
Paper summary: FlowRL: Matching Reward Distributions for LLM Reasoning
FlowRL replaces reward maximisation with matching the whole reward distribution (via a learnable partition function and reverse KL), to keep diverse reasoning paths alive; it reports +10.0% over GRPO on maths.
26 Aug 2025 · Paper summary
Paper summary: InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency
InternVL3.5 adds Cascade RL (offline then online RL), a Visual Resolution Router and decoupled vision/language deployment, reporting up to +16% reasoning and 4.05× faster inference than InternVL3.
11 Aug 2025 · Paper summary
Paper summary: GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models
GLM-4.5 is an open MoE LLM (355B total, 32B active) with hybrid thinking and direct modes, trained on 23T tokens plus expert iteration and RL, reporting 70.1% TAU-Bench, 91.0% AIME 24 and 64.2% SWE-bench Verified.
2 Jul 2025 · Paper summary
Paper summary: GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning
GLM-4.1V-Thinking and GLM-4.5V use a reasoning-centric training framework with Reinforcement Learning with Curriculum Sampling (RLCS); GLM-4.5V is reported to lead open models of its size on nearly all of 42 benchmarks.
17 Jun 2025 · Paper summary
Paper summary: MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention
MiniMax-M1 is an open-weight hybrid-attention reasoning model (456B total, 45.9B active, 1M-token context) trained with large-scale RL and a new algorithm, CISPO, at a reported rental cost of $534,700.
19 May 2025 · Paper summary
Paper summary: Qwen3 Technical Report
Qwen3 merges a "thinking" mode for multi-step reasoning and a fast non-thinking mode in one model family (0.6B to 235B, dense and MoE), adds a thinking budget, and expands to 119 languages.
7 May 2025 · Paper summary
Paper summary: Absolute Zero: Reinforced Self-play Reasoning with Zero Data
Absolute Zero trains a model to propose its own coding and maths-style tasks and then solve them, with a code executor checking both, so reasoning improves through self-play with no external training data.
30 Apr 2025 · Paper summary
Paper summary: Reinforcement Learning for Reasoning in Large Language Models with One Training Example
RL with verifiable rewards on a single maths training example lifts Qwen2.5-Math-1.5B on MATH500 from 36.0% to 73.6%, matching training on a 1.2k-example dataset.
21 Apr 2025 · Paper summary
Paper summary: Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?
Measuring pass@k at large k, this study finds RL with verifiable rewards makes models better at k = 1 but doesn't expand what the base model can solve at all; distillation, by contrast, can.
7 Mar 2025 · Paper summary
Paper summary: START: Self-taught Reasoner with Tools
START teaches a long chain-of-thought reasoning model to call a Python interpreter mid-reasoning, using inserted hints ("Wait, maybe using Python here is a good idea") and rejection-sampling fine-tuning.
10 Feb 2025 · Paper summary
Paper summary: Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
A language model that "thinks" by looping a recurrent block more times at test time, reasoning in latent space instead of writing out more chain-of-thought tokens; a 3.5B proof of concept.
28 Jan 2025 · AI updates
Open-R1: Hugging Face sets out to reproduce DeepSeek-R1 openly
A week after DeepSeek-R1, Hugging Face launched Open-R1 to reconstruct its data and training pipeline in the open, since DeepSeek released weights and a report but no training code or datasets.
23 Jan 2025 · Paper summary
Paper summary: DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
DeepSeek-R1 shows that large-scale reinforcement learning, without human-labelled reasoning traces, can bring out advanced reasoning such as self-reflection and verification in a pretrained LLM.
9 Jan 2025 · Paper summary
Paper summary: rStar-Math: Small LLMs Can Master Math Reasoning with Self-Evolved Deep Thinking
rStar-Math gets small language models to strong maths reasoning with Monte Carlo Tree Search guided by a small process preference model, both improved through self-evolution without distillation from stronger models.
22 Nov 2024 · Paper summary
Paper summary: Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization
Mixed Preference Optimization and the automatically built MMPR dataset improve chain-of-thought reasoning in multimodal LLMs (InternVL2-8B and 76B) beyond what supervised fine-tuning alone gives.
17 Nov 2024 · Paper summary
Paper summary: LLaVA-CoT: Let Vision Language Models Reason Step-by-Step
LLaVA-CoT trains a vision-language model to reason in explicit stages (summary, visual interpretation, logical reasoning, conclusion) and adds a stage-wise retracing search at test time.
16 Feb 2024 · Paper summary
Paper summary: Chain-of-Thought Reasoning Without Prompting
Chain-of-thought reasoning can appear without any prompting: branching on the top-k first tokens during decoding often reveals reasoning paths, and those paths come with higher answer confidence.
6 Feb 2024 · Paper summary
Paper summary: DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
DeepSeekMath 7B continues pre-training on 120B maths tokens mined from Common Crawl and introduces GRPO, a PPO variant without a value network, reaching 51.7% on the MATH benchmark.