Kaggle
PulseAugur coverage of Kaggle — every cluster mentioning Kaggle across labs, papers, and developer communities, ranked by signal.
- subsidiary of Google 100%
- subsidiary of .google 100%
- subsidiary of Alphabet Inc. 100%
- developed by Gemma 2 95%
- used by transformers 70%
- used by SGLang 70%
- used by llama.cpp 70%
- used by llama-cpp-python 70%
- affiliated with Eugene Yan 70%
- instance of Gemma 2 70%
- used by SK Telecom 70%
- used by Google Colab 60%
- 2026-07-29 regulatory Kaggle removed a stroke dataset due to copyright infringement concerns. source
- 2026-07-17 controversy A $25,000 Google DeepMind Kaggle Grand Prize was awarded to content identified as 'AI slop' due to a prompt injection attack, raising concerns about AI competition integrity. source
17 day(s) with sentiment data
-
New AI model efficiently classifies tomato diseases with minimal parameters
Researchers have developed a new AI model called CoAtNet-DeepMoE, designed for efficient tomato disease classification. This hybrid architecture combines convolutional and attention mechanisms with a DeepSeek Mixture of…
-
Ternary Bonsai 2 27B model now available for local use
The prism-ml/Ternary-Bonsai-2-27B-gguf model is now available for use with various local applications and inference providers. Instructions are provided for integrating the model with tools such as llama.cpp, vLLM, Olla…
-
XingChen-AGI releases Xing4.0-29B-A4B with 256K context length
XingChen-AGI has released Xing4.0-29B-A4B, a new large language model in the Xing series, formerly known as TeleChat. This model boasts 29 billion parameters with only 4 billion activated per token, enabling a native co…
-
Developer fine-tunes 7B medical AI for free using Kaggle GPUs
A developer details a cost-effective method for fine-tuning a 7-billion parameter medical AI model using free Kaggle GPUs. The process involved techniques like QLoRA and FAISS RAG, along with custom safety measures, to …
-
Developer open-sources low-cost tool to prevent LLM data poisoning
A developer has created and open-sourced a lightweight tool called Beatriz Epistemic Gate to combat data poisoning during the fine-tuning of large language models. This tool acts as a proxy, verifying generated text aga…
-
Kaggle winners leverage GPT5.4, Gemini3.1, ClaudeOpus4.6 in AI-assisted coding
A Kaggle competition winner's write-up highlights a significant shift in AI-assisted coding, where frontier models like GPT5.4, Gemini3.1, and ClaudeOpus4.6 are now prominently featured in winning strategies. The author…
-
Agnes-AI releases 33B multimodal model with 262K context window
Agnes-AI has released Agnes-3.0-Flash, a 33 billion parameter multimodal model with a 262,144 token context window. The model features a hybrid-attention architecture, combining gated delta-rule recurrent layers with st…
-
Code quality and ML performance largely unrelated, Kaggle study finds
A large-scale empirical study of 265,363 Python notebooks from Kaggle competitions revealed that general Python code quality is largely decoupled from machine learning performance. However, violations of ML-specific cod…
-
Comfy-Org/YuE2 Model Available on Hugging Face with Integration Guides
The Comfy-Org/YuE2 model is now available on Hugging Face, with instructions provided for its integration into various platforms. Users can find guidance for using the model with libraries, inference providers, notebook…
-
DeepSeek-V4.1-Flash model released with safety guardrails removed
The dealignai team has released an uncensored version of the DeepSeek-V4.1-Flash model, named DeepSeek-V4.1-Flash-UNCENSORED-FP8. This version features proprietary weight-level abliteration, surgically removing safety g…
-
Yandex releases AliceAI-T5-35B-A0.6B with sparse MoE layers
Yandex has released the AliceAI-T5-35B-A0.6B model, a language model featuring an encoder-decoder architecture with sparse Mixture-of-Experts (MoE) layers. This model boasts 34.35 billion unique parameters and utilizes …
-
DeepSeek releases V4.1 Flash with efficient MoE architecture
DeepSeek has officially released its V4.1 Flash model, a 552 billion parameter Mixture-of-Experts (MoE) model featuring a Causal-Encoder-Decoder (CED) architecture and native multimodal capabilities. This new model is d…
-
DIY AI researcher develops $0 epistemic gate to combat LLM manipulation
An independent researcher details a $0 project to develop an epistemic gate for large language models, aiming to prevent manipulation and ensure factual accuracy. Facing hardware limitations, the researcher conducted 16…
-
Edge0 releases 35B MoE LLM for low-memory devices
Edge0 has released a preview of its Edge0-35B-A3B model, a 35 billion parameter sparse Mixture-of-Experts (MoE) large language model designed to run efficiently on devices with limited memory. The model requires under 3…
-
Hugging Face hosts new multimodal Qwen models with broad integration support
The ukisai/Swift-Qwen3.8-27B-GGUF and ukisai/Swift-Qwen3.8-27b models are now available on Hugging Face, offering multimodal capabilities. These models can be integrated with various libraries and inference providers, i…
-
Nex AGI unveils Nex-N2.5 agentic model family with multimodal and trillion-parameter options
Nex AGI has released its new family of agentic models, Nex-N2.5, designed for long-horizon tasks and real-world environments. The models are available in three sizes: mini, Pro, and Max. The mini and Pro versions build …
-
Qwen3.8-Flash-Next model released with GSQ-RCO quantization, reducing size and boosting speed
A new version of the Qwen3.8-Flash-Next model, named GSQ-RCO, has been released, offering significantly reduced file sizes while maintaining near-baseline quality. This quantization method cuts the model size from appro…
-
User deploys Qwen3.8 model via llama.cpp on Kaggle
A user has successfully set up the llama.cpp software on Kaggle, enabling access from their personal computer via Cloudflare Tunnel. They are utilizing the Qwen3.8 27B+ MTP model, achieving a performance of 19 tokens pe…
-
TokenRhythm releases NeoHorse-1-9B for recursive self-improvement
TokenRhythm has released NeoHorse-1-9B, a 9-billion parameter causal language model. This model is an initial prototype for recursive self-improvement, built upon Qwen3.5-9B and fine-tuned for agentic tasks, tool use, c…
-
OpenBMB releases MiniCPM5-2B, a 2.52B model for on-device use
OpenBMB has released MiniCPM5-2B, a 2.52 billion parameter dense language model designed for on-device deployment. This model achieves state-of-the-art performance among open-source models of its size, excelling in codi…