uz
Feedback

Firibgarlarga uchmang! Telemetrio bunday kanallarni topadi va belgilaydi πŸ‘‰ Belgini koβ€˜rmoqchi boβ€˜lsangiz, obuna boβ€˜ling

Data science research papers

Data science research papers

Kanalga Telegram’da oβ€˜tish

Machine learning and data science research papers Key ML and AI papers with code and GitHub repos. Simple way to follow current research. Join πŸ‘‰ https://rebrand.ly/bigdatachannels DMCA: @disclosure_bds Contact: @mldatascientist

Ko'proq ko'rsatish
3 242
Obunachilar
+124 soatlar
+107 kun
+1730 kun
Postlar arxiv
Scaling Automatic Research Agents via World Models πŸ“… Publication Date: Aug 29, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2608.125
Scaling Automatic Research Agents via World Models πŸ“… Publication Date: Aug 29, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2608.12564.pdf πŸ”— Code: https://github.com/xiyuanyang45/WMRL πŸ’» Project Page: https://xiyuanyang45.github.io/WMRL/ πŸ“ Description: World Model RL replaces costly environment execution with a learned world model and applies debiasing and denoising to accelerate post-training of autonomous research agents. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness πŸ“… Publication Date: Sep 08, 20
NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness πŸ“… Publication Date: Sep 08, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.08183.pdf πŸ”— Code: https://github.com/TokenRhythm/NeoHorse πŸ’» Project Page: https://huggingface.co/collections/TokenRhythm/neohorse-1 πŸ“ Description: NeoHorse-1 uses agentic post-training with intelligent routing, structured feedback loops, and curriculum-based distillation to improve model capabilities across agent benchmarks. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

Dr. Claw: An AI Scientist Workspace for Vibe Research πŸ“… Publication Date: Aug 31, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.
Dr. Claw: An AI Scientist Workspace for Vibe Research πŸ“… Publication Date: Aug 31, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.00365.pdf πŸ”— Code: https://github.com/OpenLAIR/dr-claw πŸ’» Project Page: https://github.com/OpenLAIR/dr-claw/ πŸ“ Description: Command-line coding agents (e.g., Claude Code, Gemini CLI) can already read and write files and sustain long sessions, yet end-to-end research still fragments across chat tools, IDEs, terminals, and writing environments, and the decisions that make it auditable are rarely preserved. We present Dr. C... βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

Unlocking Lossless Speedups in LLMs via Discrete Diffusion πŸ“… Publication Date: Sep 03, 2026 πŸ“„ Paper: https://arxiv.org/pdf/
Unlocking Lossless Speedups in LLMs via Discrete Diffusion πŸ“… Publication Date: Sep 03, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.04010.pdf πŸ”— Code: https://github.com/ifm-ai/uno πŸ’» Project Page: https://s-sahoo.com/uno/ πŸ“ Description: Diffusion-augmented autoregressive language models use parallel token sampling via distilled diffusion weights and a specialized sampler to accelerate inference without quality loss or draft models. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Pap
Bilevel Coordinated Reflection: A Game-Theoretic Approach to Multi-Agent LLM Systems πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.02750.pdf πŸ”— Code: https://github.com/YihangChen9/Bilevel-Coordinated-Reflection πŸ“ Description: The study formalizes multi-agent LLM coordination via bilevel games and stochastic memory reflection, introducing a grounded evaluation gate and SRMA algorithm with convergence guarantees, validated on SWE-bench. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

πŸ“’ Advertising in this channel You can place an ad via Telegaβ€€io. It takes just a few minutes. Formats and current rates: Vie
πŸ“’ Advertising in this channel You can place an ad via Telegaβ€€io. It takes just a few minutes. Formats and current rates: View details

Show-Harness: Just a VLM Agent Can Play Robots πŸ“… Publication Date: Sep 09, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.10522.p
Show-Harness: Just a VLM Agent Can Play Robots πŸ“… Publication Date: Sep 09, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.10522.pdf πŸ”— Code: https://github.com/showlab/Show-Harness πŸ’» Project Page: https://showlab.github.io/Show-Harness/ πŸ“ Description: Show-Harness links vision-language models to robot control via discrete semantic actions interpreted by embodiment-specific modules, enabling zero-shot and efficient fine-tuned deployment across robots and GUIs. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

Omni Interaction Agent Technical Report πŸ“… Publication Date: Sep 08, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.08977.pdf πŸ”— C
Omni Interaction Agent Technical Report πŸ“… Publication Date: Sep 08, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.08977.pdf πŸ”— Code: https://github.com/Omni-Interaction-Gander/Omni-Interaction-Agent πŸ’» Project Page: https://omni-interaction-gander.github.io/Omni-Interaction-Agent/ πŸ“ Description: Gander is an end-to-end framework that integrates continuous multi-modal streaming, real-time full-duplex interaction, and agentic reasoning through a Cerebellum-Brain architecture and a chunk-level token stream design. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

SpatialBlock: Enhancing Spatial Intelligence in LVLMs via Synthetic Block-Stacking Problem πŸ“… Publication Date: Sep 07, 2026
SpatialBlock: Enhancing Spatial Intelligence in LVLMs via Synthetic Block-Stacking Problem πŸ“… Publication Date: Sep 07, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.07064.pdf πŸ”— Code: https://github.com/rsoohyun/SpatialBlock πŸ“ Description: Large vision-language models trained on synthetic block-manipulation tasks improve 3D spatial reasoning and generalize to real-world visual tasks. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

Programmable World Model πŸ“… Publication Date: Sep 09, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.10540.pdf πŸ”— Code: https://gi
Programmable World Model πŸ“… Publication Date: Sep 09, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.10540.pdf πŸ”— Code: https://github.com/AlayaLab/pwm πŸ’» Project Page: https://alaya-lab.github.io/pwm/ πŸ“ Description: A programmable world model separates explicit state evolution from video generation using executable rules and 3D bounding boxes to maintain persistent, controllable environments. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Paper: https://arxiv.org
Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.02749.pdf πŸ”— Code: https://github.com/VectorSpaceLab/AREX-Skill πŸ“ Description: DisCo is a research agent that distills operational knowledge into reusable skills, significantly improving autonomous ML research performance across benchmarks. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

Compile by Training: Turning Natural-Language Specifications into Local Neural Functions πŸ“… Publication Date: Sep 03, 2026 πŸ“„
Compile by Training: Turning Natural-Language Specifications into Local Neural Functions πŸ“… Publication Date: Sep 03, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.04199.pdf πŸ”— Code: https://github.com/programasweights/compile-by-training πŸ’» Project Page: https://programasweights.com/playground?compiler=paw-ft-bs48 πŸ“ Description: Compile by training converts natural-language specifications into reusable neural functions by distilling teacher-generated examples into small adapters, enabling efficient deployment without remote model dependencies. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Paper: http
SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.02886.pdf πŸ”— Code: https://github.com/Junchao-cs/SolarWM πŸ’» Project Page: https://junchao-cs.github.io/SolarWM-Web/ πŸ“ Description: SolarWM provides an open framework and unified training recipe for building interactive video world models across diverse data sources and generator backbones, enabling long-horizon real-time rollouts. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

RoboTok: An Internet-Scale Data Engine for Human Demonstration Retrieval and Dexterous Manipulation Learning πŸ“… Publication D
RoboTok: An Internet-Scale Data Engine for Human Demonstration Retrieval and Dexterous Manipulation Learning πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.03199.pdf πŸ”— Code: https://github.com/Rice-RobotPI-Lab/RoboTok-Code πŸ’» Project Page: https://rice-robotpi-lab.github.io/RoboTok/ πŸ“ Description: RoboTok retrieves relevant human manipulation videos from the web using a latent motion space derived from 3D hand trajectories to improve robot policy training. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Paper: https://arxiv.or
EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction πŸ“… Publication Date: Sep 02, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.02783.pdf πŸ”— Code: https://github.com/inphotoo/earlyeval πŸ“ Description: EarlyEval predicts agent outcomes from intermediate behavior to reduce evaluation cost by halting runs early with minimal accuracy loss. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

RRSI: Regularized Recursive Self-Improvement of Agent Harnesses πŸ“… Publication Date: Sep 21, 2026 πŸ“„ Paper: https://arxiv.org
RRSI: Regularized Recursive Self-Improvement of Agent Harnesses πŸ“… Publication Date: Sep 21, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.24972.pdf πŸ”— Code: https://github.com/google-research/rrsi πŸ’» Project Page: https://regularized-rsi.com/ πŸ“ Description: An LLM agent's capability is largely magnified by its harness, namely the prompts, control flow, tooling, memory, and context management surrounding the frozen backbone model. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

LatentPress: Context Compression Beyond Text and Vision πŸ“… Publication Date: Sep 01, 2026 πŸ“„ Paper: https://arxiv.org/pdf/260
LatentPress: Context Compression Beyond Text and Vision πŸ“… Publication Date: Sep 01, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2609.01507.pdf πŸ”— Code: https://github.com/HJSang/LatentPress πŸ’» Project Page: https://github.com/HJSang/LatentPress πŸ“ Description: LatentPress compresses conversational and document context into continuous memory tokens read directly by a frozen decoder, achieving high compression with faster inference and improved accuracy over text or OCR methods. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering πŸ“… Publication Date: Aug 28, 2026 πŸ“„ Paper: https:
LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering πŸ“… Publication Date: Aug 28, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2608.28281.pdf πŸ”— Code: https://github.com/AMAP-ML/LoopArena πŸ’» Project Page: https://amap-ml.github.io/LoopArena/ πŸ“ Description: LoopArena benchmarks how well a controller model guides a separate coding agent through long tasks, revealing low strict success rates and significant cost reductions. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

UrbanGround: From Local Perception to Spatial Agency in a Real-Scale City πŸ“… Publication Date: Aug 27, 2026 πŸ“„ Paper: https:/
UrbanGround: From Local Perception to Spatial Agency in a Real-Scale City πŸ“… Publication Date: Aug 27, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2608.27456.pdf πŸ”— Code: https://github.com/UrbanGround/UrbanGround πŸ’» Project Page: https://urbanground.github.io πŸ“ Description: UrbanGround evaluates whether multimodal language model agents can sustain reliable navigation and spatial reasoning in a realistic 3D city replica, revealing that local perceptual skills fail to compose into extended goal-directed behavior. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community

PAWBench: How Far Are We from Probabilistically Aligned World Modeling? πŸ“… Publication Date: Aug 27, 2026 πŸ“„ Paper: https://a
PAWBench: How Far Are We from Probabilistically Aligned World Modeling? πŸ“… Publication Date: Aug 27, 2026 πŸ“„ Paper: https://arxiv.org/pdf/2608.27345.pdf πŸ”— Code: https://github.com/Andrew0613/PAWBench πŸ’» Project Page: https://pawbench.github.io/ πŸ“ Description: The study formalizes probabilistic alignment for world models, introduces PAWBench and PAWEval to evaluate video generators as stochastic samplers, and finds current models fail to match reference behavior distributions. βž–βž–βž–βž–βž–βž–βž–βž–βž–βž–βž– πŸ“„ More research papers β†’ @data_science_research_papers Part of @bigdataspecialist community