https://magic-with-latents.github.io/latent/posts/flow_models/part1/vector_fields.html
Heuristics AI
@heuristics_ai
Ai research updates
LLMs
Reinforcement learning
Deep learning
GANs
Stable diffusion
Transformers
NLP
Kindly join (☞ ಠ_ಠ)☞ @heuristics_ai
Heuristics AI (English)
Welcome to Heuristics AI, your go-to Telegram channel for all the latest updates and advancements in the world of Artificial Intelligence. Whether you're a seasoned AI expert or just starting to dip your toes into the field, this channel has something for everyone. From Ai research updates to discussions on LLMs, Reinforcement learning, Deep learning, GANs, Stable diffusion, Transformers, NLP, and more, you'll find a wealth of valuable information and insights here.
Join our community of like-minded individuals who are passionate about pushing the boundaries of AI technology forward. Stay informed about the cutting-edge developments in the industry, learn about the latest trends, and engage in stimulating conversations with fellow enthusiasts.
Don't miss out on this opportunity to expand your knowledge and stay ahead of the curve in the exciting world of Artificial Intelligence. Join Heuristics AI today and be part of a dynamic community that is shaping the future of AI. Kindly join (☞ ಠ_ಠ)☞ @heuristics_ai
Heuristics AI
18 Feb, 13:31
Grok 3 got released today by XAI . It's available for all premium+ users of x.com
Heuristics AI
18 Feb, 13:31
Paper page - DarwinLM: Evolutionary Structured Pruning of Large Language Models
https://huggingface.co/papers/2502.07780
https://huggingface.co/papers/2502.07780
Heuristics AI
18 Feb, 02:29
Watch R1 "think" with animated chains of thought https://github.com/dhealy05/frames_of_mind
Heuristics AI
18 Feb, 02:29
I helped fix sleep-wake hangs on Linux with AMD GPUs https://nyanpasu64.gitlab.io/blog/amdgpu-sleep-wake-hang/
Heuristics AI
18 Feb, 02:29
George Eliot's reflections on AGI from 1879 https://www.learningfromexamples.com/p/reflections-on-superintelligence
Heuristics AI
17 Feb, 13:36
open-thoughts/OpenThinker-32B · Hugging Face
https://huggingface.co/open-thoughts/OpenThinker-32B
https://huggingface.co/open-thoughts/OpenThinker-32B
Heuristics AI
16 Feb, 12:09
https://davidstutz.de/how-i-prepared-for-deepmind-and-google-ai-research-internship-interviews-in-2019/
Heuristics AI
16 Feb, 09:53
Paper page - Competitive Programming with Large Reasoning Models
https://huggingface.co/papers/2502.06807
https://huggingface.co/papers/2502.06807
Heuristics AI
16 Feb, 05:49
https://hsf-training.github.io/hsf-training-ml-gpu-webpage/aio/index.html
Heuristics AI
16 Feb, 05:46
https://predibase.com/blog/teaching-ai-to-write-gpu-code-a-deep-dive-into-reinforcement-fine-tuning
Heuristics AI
10 Feb, 15:23
PhD Knowledge Not Required: A Reasoning Challenge for Large Language Models https://arxiv.org/abs/2502.01584
Heuristics AI
10 Feb, 15:22
TL;DR of Deep Dive into LLMs Like ChatGPT by Andrej Karpathy https://anfalmushtaq.com/articles/deep-dive-into-llms-like-chatgpt-tldr
Heuristics AI
10 Feb, 15:21
The Anthropic Economic Index https://www.anthropic.com/news/the-anthropic-economic-index
Heuristics AI
10 Feb, 15:18
Paper page - Step Back to Leap Forward: Self-Backtracking for Boosting Reasoning of Language Models
https://huggingface.co/papers/2502.04404
https://huggingface.co/papers/2502.04404
Heuristics AI
06 Feb, 13:44
https://rbcborealis.com/research-blogs/odes-and-sdes-for-machine-learning/
Heuristics AI
06 Feb, 05:29
Incredible opportunity to do great mechanistic interpretability research under the awesome guidance of
Neel Nanda
Application: https://tinyurl.com/neel-mats-app
Twitter thread: https://x.com/NeelNanda5/status/1887274059408548208
Neel Nanda
Application: https://tinyurl.com/neel-mats-app
Twitter thread: https://x.com/NeelNanda5/status/1887274059408548208
Heuristics AI
03 Feb, 16:45
DeepSeek-R1 more readily generates dangerous content than other large language models | Computer Weekly
https://www.computerweekly.com/news/366618734/DeepSeek-R1-more-readily-generates-dangerous-content-than-other-large-language-models#:~:text=Turning%20to%20cyber%20security%20risks,malware%2C%20Trojans%20and%20other%20exploits.
https://www.computerweekly.com/news/366618734/DeepSeek-R1-more-readily-generates-dangerous-content-than-other-large-language-models#:~:text=Turning%20to%20cyber%20security%20risks,malware%2C%20Trojans%20and%20other%20exploits.
Heuristics AI
03 Feb, 13:18
Anthropic: "Applicants should not use AI assistants" https://simonwillison.net/2025/Feb/2/anthropic/
Heuristics AI
01 Feb, 03:52
o3-mini is now available to all users: https://openai.com/index/openai-o3-mini/
Heuristics AI
30 Jan, 10:00
Qwen2.5-Max: Exploring the Intelligence of Large-scale MoE Model | Qwen
https://qwenlm.github.io/blog/qwen2.5-max/
https://qwenlm.github.io/blog/qwen2.5-max/
Heuristics AI
29 Jan, 17:31
Introducing ChatGPT Gov | OpenAI
https://openai.com/global-affairs/introducing-chatgpt-gov/
https://openai.com/global-affairs/introducing-chatgpt-gov/
Heuristics AI
29 Jan, 15:26
OpenAI Furious DeepSeek Might Have Stolen All the Data OpenAI Stole from Us https://www.404media.co/openai-furious-deepseek-might-have-stolen-all-the-data-openai-stole-from-us/
Heuristics AI
29 Jan, 05:38
Run DeepSeek R1 Dynamic 1.58-bit https://unsloth.ai/blog/deepseekr1-dynamic
Heuristics AI
29 Jan, 05:38
Promising results from DeepSeek R1 for code https://simonwillison.net/2025/Jan/27/llamacpp-pr/
Heuristics AI
29 Jan, 05:38
Janus Pro 1B running 100% locally in-browser on WebGPU https://old.reddit.com/r/LocalLLaMA/comments/1ibnso0/janus_pro_1b_running_100_locally_inbrowser_on/
Heuristics AI
29 Jan, 05:38
Berkeley Researchers Replicate DeepSeek R1's Core Tech for Just $30: A Small Mod https://xyzlabs.substack.com/p/berkeley-researchers-replicate-deepseek
Heuristics AI
29 Jan, 05:38
Sam Altman said startups with $10M were 'hopeless' competing with OpenAI https://www.tomshardware.com/tech-industry/artificial-intelligence/sam-altman-said-startups-with-only-usd10-million-were-totally-hopeless-competing-with-openai-deepseeks-disruption-says-otherwise
Heuristics AI
29 Jan, 05:38
Qwen2.5-Max: Exploring the intelligence of large-scale MoE model https://qwenlm.github.io/blog/qwen2.5-max/
Heuristics AI
29 Jan, 05:38
DeepSeek's multi-head latent attention and other KV cache tricks https://www.pyspur.dev/blog/multi-head-latent-attention-kv-cache-paper-list
Heuristics AI
28 Jan, 09:38
Sutton RL Reading List https://docs.google.com/document/d/1juudZLXpqMsuAXg7zGFlkRdBf8hffDzSChWkHAmJci0/edit?tab=t.0#heading=h.rsj4tqabnrgw
Heuristics AI
28 Jan, 06:26
The Illustrated DeepSeek-R1 https://newsletter.languagemodels.co/p/the-illustrated-deepseek-r1
Heuristics AI
28 Jan, 06:26
Nvidia sheds almost $600B in market cap, biggest one-day loss in US history https://www.cnbc.com/2025/01/27/nvidia-sheds-almost-600-billion-in-market-cap-biggest-drop-ever.html
Heuristics AI
28 Jan, 06:26
I trusted an LLM, now I'm on day 4 of an afternoon project https://nemo.foo/blog/day-4-of-an-afternoon-project
Heuristics AI
28 Jan, 06:26
Nvidia’s $589B DeepSeek rout https://finance.yahoo.com/news/asml-sinks-china-ai-startup-081823609.html
Heuristics AI
28 Jan, 06:26
DeepSeek releases Janus Pro, a text-to-image generator [pdf] https://github.com/deepseek-ai/Janus/blob/main/janus_pro_tech_report.pdf
Heuristics AI
28 Jan, 06:23
Using AI for Coding: My Journey with Cline and LLMs https://pgaleone.eu/ai/coding/2025/01/26/using-ai-for-coding-my-experience/
Heuristics AI
22 Jan, 11:31
Metacognitive laziness: Effects of generative AI on learning motivation https://bera-journals.onlinelibrary.wiley.com/doi/10.1111/bjet.13544
Heuristics AI
22 Jan, 11:31
Kimi K1.5: Scaling Reinforcement Learning with LLMs https://github.com/MoonshotAI/Kimi-k1.5
Heuristics AI
22 Jan, 11:31
Should we use AI and LLMs for Christian apologetics? (2024) https://lukeplant.me.uk/blog/posts/should-we-use-llms-for-christian-apologetics/
Heuristics AI
22 Jan, 11:31
Hunyuan3D 2.0 – High-Resolution 3D Assets Generation https://github.com/Tencent/Hunyuan3D-2
Heuristics AI
22 Jan, 11:31
Stargate Project: SoftBank, OpenAI, Oracle, MGX to build data centers https://apnews.com/article/trump-ai-openai-oracle-softbank-son-altman-ellison-be261f8a8ee07a0623d4170397348c41
Heuristics AI
22 Jan, 11:31
Tensor Product Attention Is All You Need https://arxiv.org/abs/2501.06425
Heuristics AI
22 Jan, 11:31
Official DeepSeek R1 Now on Ollama https://ollama.com/library/deepseek-r1
Heuristics AI
22 Jan, 10:52
O1 Replication Journey: A Strategic Progress Report -- Part 1
https://arxiv.org/abs/2410.18982
https://github.com/GAIR-NLP/O1-Journey
O1 Replication Journey -- Part 2: Surpassing O1-preview through Simple Distillation, Big Progress or Bitter Lesson?
https://arxiv.org/abs/2411.16489
O1 Replication Journey -- Part 3: Inference-time Scaling for Medical Reasoning
https://arxiv.org/abs/2501.06458
https://arxiv.org/abs/2410.18982
https://github.com/GAIR-NLP/O1-Journey
O1 Replication Journey -- Part 2: Surpassing O1-preview through Simple Distillation, Big Progress or Bitter Lesson?
https://arxiv.org/abs/2411.16489
O1 Replication Journey -- Part 3: Inference-time Scaling for Medical Reasoning
https://arxiv.org/abs/2501.06458
Heuristics AI
22 Jan, 10:52
Advancing Language Model Reasoning through Reinforcement Learning and Inference Scaling
https://arxiv.org/abs/2501.11651
https://github.com/THUDM/T1
https://arxiv.org/abs/2501.11651
https://github.com/THUDM/T1
Heuristics AI
22 Jan, 10:52
Reasoning Language Models: A Blueprint
https://arxiv.org/abs/2501.11223
https://arxiv.org/abs/2501.11223
Heuristics AI
22 Jan, 10:52
RedStar: Does Scaling Long-CoT Data Unlock Better Slow-Reasoning Systems?
https://arxiv.org/abs/2501.11284
https://huggingface.co/RedStar-Reasoning
https://arxiv.org/abs/2501.11284
https://huggingface.co/RedStar-Reasoning
Heuristics AI
21 Jan, 10:54
KIMI K1.5:
SCALING REINFORCEMENT LEARNING WITH LLMS
TECHNICAL REPORT OF KIMI K1.5
SCALING REINFORCEMENT LEARNING WITH LLMS
TECHNICAL REPORT OF KIMI K1.5
Heuristics AI
21 Jan, 08:39
⚡ 🚀 Top 10 Trending GitHub Repositories, January 2025 🔥 - DEV Community
https://dev.to/leamsigc/top-10-trending-github-repositories-january-2025-1f24
https://dev.to/leamsigc/top-10-trending-github-repositories-january-2025-1f24
Heuristics AI
13 Jan, 09:27
https://www.abrarsheikh.dev/content/head_attention.html popular attention mechanisms along with code and their performance benchmarks,
Heuristics AI
10 Jan, 17:05
Resources to get started with Triton: https://github.com/rkinas/triton-resources
Heuristics AI
09 Jan, 00:42
Medical large language models are vulnerable to data-poisoning attacks
https://www.nature.com/articles/s41591-024-03445-1
https://www.nature.com/articles/s41591-024-03445-1
Heuristics AI
08 Jan, 09:00
Paper page - VITA-1.5: Towards GPT-4o Level Real-Time Vision and Speech Interaction
https://huggingface.co/papers/2501.01957
https://huggingface.co/papers/2501.01957
Heuristics AI
08 Jan, 08:57
Paper page - REINFORCE++: A Simple and Efficient Approach for Aligning Large Language Models
https://huggingface.co/papers/2501.03262
https://huggingface.co/papers/2501.03262
Heuristics AI
07 Jan, 17:09
DeepFace: A lightweight deep face recognition library for Python https://github.com/serengil/deepface
Heuristics AI
07 Jan, 17:09
Time-Series Anomaly Detection: A Decade Review https://arxiv.org/abs/2412.20512
Heuristics AI
07 Jan, 17:09
Used Meta AI, now Instagram is using my face on ads targeted at me https://old.reddit.com/r/ABoringDystopia/comments/1ht7fft/used_meta_ai_to_edit_a_selfie_now_instagram_is/
Heuristics AI
07 Jan, 17:09
How I program with LLMs https://crawshaw.io/blog/programming-with-llms
Heuristics AI
07 Jan, 17:09
LLMs and Code Optimization https://wiredream.com/llm-optimizing-digit-diff/
Heuristics AI
07 Jan, 17:09
Nvidia's Project Digits is a 'personal AI supercomputer' https://techcrunch.com/2025/01/06/nvidias-project-digits-is-a-personal-ai-computer/
Heuristics AI
07 Jan, 17:09
Nvidia announces $3k personal AI supercomputer called Digits https://www.theverge.com/2025/1/6/24337530/nvidia-ces-digits-super-computer-ai
Heuristics AI
05 Jan, 00:34
#tools
#MLSecOps
LLM Vulnerability Scanner
https://github.com/NVIDIA/garak
]-> Framework for Security Probing LLMs (.pdf)
#MLSecOps
LLM Vulnerability Scanner
https://github.com/NVIDIA/garak
]-> Framework for Security Probing LLMs (.pdf)
Heuristics AI
05 Jan, 00:34
#Research
"Hacking CTFs with Plain Agents", 2024.
]-> https://github.com/palisaderesearch/intercode
"Hacking CTFs with Plain Agents", 2024.
]-> https://github.com/palisaderesearch/intercode
Heuristics AI
05 Jan, 00:32
#MLSecOps
"CySecBench: Generative AI-based CyberSecurity-focused Prompt Dataset for Benchmarking Large Language Models", 2025.
"CySecBench: Generative AI-based CyberSecurity-focused Prompt Dataset for Benchmarking Large Language Models", 2025.
Heuristics AI
03 Jan, 00:07
We fine-tuned Llama and got 4.2x Sonnet 3.5 accuracy for code generation https://finecodex.com/
Heuristics AI
03 Jan, 00:07
Notes on the New Deepseek v3 https://composio.dev/blog/notes-on-new-deepseek-v3/
Heuristics AI
03 Jan, 00:07
TinyStories: How Small Can Language Models Be and Still Speak Coherent English? (2023) https://arxiv.org/abs/2305.07759
Heuristics AI
03 Jan, 00:07
Building a Knowledge System That Enhances Rather Than Replaces Thought https://nsavage.substack.com/p/beyond-rag-building-a-knowledge-management
Heuristics AI
03 Jan, 00:07
Things we learned about LLMs in 2024 https://simonwillison.net/2024/Dec/31/llms-in-2024/
Heuristics AI
03 Jan, 00:07
Developing inside a virtual machine https://blog.disintegrator.dev/posts/dev-virtual-machine/
Heuristics AI
03 Jan, 00:07
30% drop in O1-preview accuracy when Putnam problems are slightly variated https://openreview.net/forum?id=YXnwlZe0yf¬eId=yrsGpHd0Sf
Heuristics AI
03 Jan, 00:07
RT-2: Vision-Language-Action Models (2023) https://robotics-transformer2.github.io/
Heuristics AI
03 Jan, 00:07
How AI is unlocking ancient texts https://www.nature.com/articles/d41586-024-04161-z
Heuristics AI
03 Jan, 00:07
The GPU, not the TPM, is the root of hardware DRM https://mjg59.dreamwidth.org/70954.html
Heuristics AI
03 Jan, 00:07
Meta Wants More AI Bots on Facebook and Instagram https://nymag.com/intelligencer/article/meta-wants-more-ai-bots-on-facebook-and-instagram.html
Heuristics AI
03 Jan, 00:07
My 25-year adventure in AI and ML https://austinhenley.com/blog/25yearsofai.html
Heuristics AI
03 Jan, 00:07
Kotaemon: An open-source RAG-based tool for chatting with your documents https://github.com/Cinnamon/kotaemon
Heuristics AI
25 Dec, 18:07
Exploring LoRA – Part 1: The Idea Behind Parameter Efficient Fine-Tuning https://medium.com/inspiredbrilliance/exploring-lora-part-1-the-idea-behind-parameter-efficient-fine-tuning-and-lora-ec469d176c26
Heuristics AI
25 Dec, 18:07
I made a website to semantically search ArXiv papers https://papermatch.mitanshu.tech/
Heuristics AI
24 Dec, 07:24
Long Term Software Development https://berthub.eu/articles/posts/on-long-term-software-development/
Heuristics AI
24 Dec, 07:24
CUDA Moat Still Alive https://semianalysis.com/2024/12/22/mi300x-vs-h100-vs-h200-benchmark-part-1-training/
Heuristics AI
24 Dec, 07:24
Infinigen: Infinite Photorealistic Worlds Using Procedural Generation https://github.com/princeton-vl/infinigen
Heuristics AI
24 Dec, 07:24
Can AI do maths yet? Thoughts from a mathematician https://xenaproject.wordpress.com/2024/12/22/can-ai-do-maths-yet-thoughts-from-a-mathematician/
Heuristics AI
24 Dec, 07:24
Offline Reinforcement Learning for LLM Multi-Step Reasoning https://arxiv.org/abs/2412.16145
Heuristics AI
24 Dec, 07:24
Supernovae Evidence for Foundational Change to Cosmological Models https://arxiv.org/abs/2412.15143
Heuristics AI
24 Dec, 07:24
RAG Logger: An Open-Source Alternative to LangSmith https://github.com/Brandon-c-tech/RAG-logger
Heuristics AI
24 Dec, 07:24
Narrative Jailbreaking for Fun and Profit https://interconnected.org/home/2024/12/23/jailbreaking
Heuristics AI
24 Dec, 07:24
Making AMD GPUs competitive for LLM inference (2023) https://blog.mlc.ai/2023/08/09/Making-AMD-GPUs-competitive-for-LLM-inference
Heuristics AI
24 Dec, 07:24
AI Decodes the Calls of the Wild https://www.nature.com/immersive/d41586-024-04050-5/index.html
Heuristics AI
07 Dec, 07:17
GitHub - fishaudio/fish-speech: SOTA Open Source TTS
https://github.com/fishaudio/fish-speech
https://github.com/fishaudio/fish-speech
Heuristics AI
06 Dec, 14:49
Naturalistic Data Understanding Project
https://docs.google.com/forms/d/e/1FAIpQLSclCz67Kx62pHFqZ3C8naoLDgwnhDIV1QrY3UaWwP1MSHRC8Q/viewform?usp=send_form
https://docs.google.com/forms/d/e/1FAIpQLSclCz67Kx62pHFqZ3C8naoLDgwnhDIV1QrY3UaWwP1MSHRC8Q/viewform?usp=send_form
Heuristics AI
29 Nov, 10:18
Apply for ML Intern at Deccan AI
https://app.dover.com/apply/Deccan%20AI/e67320d5-5160-4f8c-abc5-3c6251bf103d?rs=42706078
https://app.dover.com/apply/Deccan%20AI/e67320d5-5160-4f8c-abc5-3c6251bf103d?rs=42706078
Heuristics AI
27 Nov, 14:30
GitHub - abi/screenshot-to-code: Drop in a screenshot and convert it to clean code (HTML/Tailwind/React/Vue)
https://github.com/abi/screenshot-to-code
https://github.com/abi/screenshot-to-code
Heuristics AI
26 Nov, 14:26
→ Hymba-1.5B outperforms all sub-2B public models in performance
→ Achieves 1.32% higher average accuracy than Llama-3.2-3B
→ Delivers 11.67x smaller cache size and 3.49x faster throughput compared to Llama-3.2-3B
→ Achieves 1.32% higher average accuracy than Llama-3.2-3B
→ Delivers 11.67x smaller cache size and 3.49x faster throughput compared to Llama-3.2-3B
Heuristics AI
23 Nov, 12:18
https://www.google.com/about/careers/applications/jobs/results/88049277334364870-pre-doctoral-researcher/
Heuristics AI
21 Nov, 02:56
Fastest Autograd in the West
autograd in torch
autograd in jax
autograd in python
autograd in rust
autograd in C
autograd in assembly
https://arogozhnikov.github.io/2023/12/28/fastest-autograd.html
autograd in torch
autograd in jax
autograd in python
autograd in rust
autograd in C
autograd in assembly
https://arogozhnikov.github.io/2023/12/28/fastest-autograd.html
Heuristics AI
20 Nov, 15:58
https://www.microsoft.com/en-us/research/blog/graphrag-improving-global-search-via-dynamic-community-selection/
Heuristics AI
20 Nov, 14:37
Paper page - Awaker2.5-VL: Stably Scaling MLLMs with Parameter-Efficient Mixture of Experts
https://search.app?link=https%3A%2F%2Fhuggingface.co%2Fpapers%2F2411.10669&utm_campaign=aga&utm_source=agsadl1%2Csh%2Fx%2Fgs%2Fm2%2F4
https://search.app?link=https%3A%2F%2Fhuggingface.co%2Fpapers%2F2411.10669&utm_campaign=aga&utm_source=agsadl1%2Csh%2Fx%2Fgs%2Fm2%2F4
Heuristics AI
17 Nov, 14:57
Playing Chess Against LLMs And The Mystery Of Instruct Models | Hackaday
https://search.app?link=https%3A%2F%2Fhackaday.com%2F2024%2F11%2F16%2Fplaying-chess-against-llms-and-the-mystery-of-instruct-models%2F&utm_campaign=aga&utm_source=agsadl1%2Csh%2Fx%2Fgs%2Fm2%2F4
https://search.app?link=https%3A%2F%2Fhackaday.com%2F2024%2F11%2F16%2Fplaying-chess-against-llms-and-the-mystery-of-instruct-models%2F&utm_campaign=aga&utm_source=agsadl1%2Csh%2Fx%2Fgs%2Fm2%2F4
Heuristics AI
10 Nov, 15:01
A CC-By Open-Source TTS Model with Voice Cloning https://huggingface.co/OuteAI/OuteTTS-0.1-350M
Heuristics AI
10 Nov, 15:01
LLMs have reached a point of diminishing returns https://garymarcus.substack.com/p/confirmed-llms-have-indeed-reached
Heuristics AI
10 Nov, 15:01
Physical Intelligence's first generalist policy AI can finally do your laundry https://www.physicalintelligence.company/blog/pi0
Heuristics AI
10 Nov, 15:01
When machine learning tells the wrong story https://jackcook.com/2024/11/09/bigger-fish.html
Heuristics AI
08 Nov, 13:36
AI for real-time fusion plasma behavior prediction and manipulation https://control.princeton.edu/machine-learning-for-rt-profile-control-in-tokamaks/
Heuristics AI
08 Nov, 13:36
LoRA vs. Full Fine-Tuning: An Illusion of Equivalence https://arxiv.org/abs/2410.21228
Heuristics AI
08 Nov, 13:36
Evaluating the world model implicit in a generative model https://arxiv.org/abs/2406.03689
Heuristics AI
08 Nov, 13:36
Show HN: TutoriaLLM – AI Integrated programming tutorials https://github.com/TutoriaLLM/TutoriaLLM
Heuristics AI
08 Nov, 13:36
Even Microsoft Notepad is getting AI text editing now https://www.theverge.com/2024/11/6/24289707/microsoft-notepad-ai-text-editing-rewrite
Heuristics AI
31 Oct, 07:03
enhancing intuition in diffusion
Tutorial on Diffusion Models for Imaging and Vision
what: good and comprehensive tutorial about modern diffusion models.
Diffusion is spectral autoregression
what: blogpost with python notebook.
Diffusion Models are Evolutionary Algorithms
what: another view on diffusion
- the denoising process in diffusion models can be seen as a combination of selection and mutation.
#knowledge
Tutorial on Diffusion Models for Imaging and Vision
what: good and comprehensive tutorial about modern diffusion models.
Diffusion is spectral autoregression
what: blogpost with python notebook.
Diffusion Models are Evolutionary Algorithms
what: another view on diffusion
- the denoising process in diffusion models can be seen as a combination of selection and mutation.
#knowledge
Heuristics AI
31 Oct, 06:35
LLMs know more than they show: On the intrinsic representation of hallucinations https://arxiv.org/abs/2410.02707
Heuristics AI
31 Oct, 06:35
Representing web applications as knowledge graphs https://arxiv.org/abs/2410.17258
Heuristics AI
31 Oct, 06:35
Chain-of-thought can hurt performance on tasks where thinking makes humans worse https://arxiv.org/abs/2410.21333
Heuristics AI
31 Oct, 06:34
DeepSeek v2.5 – open-source LLM comparable to GPT-4, but 95% less expensive https://www.deepseek.com/
Heuristics AI
31 Oct, 05:57
This is a great paper for understanding LLMs, especially for those of us who arent entrenched in it.
Link 🔗
@heuristics_ai
Link 🔗
@heuristics_ai
Heuristics AI
31 Oct, 04:48
https://github.com/anthropics/courses
Anthropic published a repo with courses on how to use LLMs
Anthropic published a repo with courses on how to use LLMs
Heuristics AI
31 Oct, 04:42
https://www.crackedengineers.com/
AI related jobs site by Alexa gordic
AI related jobs site by Alexa gordic
Heuristics AI
31 Oct, 04:29
AI now generates 25% of code in Google Said Sundar pichai
https://www.businessinsider.com/google-earnings-q3-2024-new-code-created-by-ai-2024-10
https://www.businessinsider.com/google-earnings-q3-2024-new-code-created-by-ai-2024-10
Heuristics AI
29 Oct, 17:41
How effective is human-AI collaboration?
A meta-analysis of 106 studies just published in Nature reports an interesting result:
On average, there was no synergy: Human–AI combinations did not perform better than both humans and AI.
In particular, when the AI alone outperformed the human alone, the human–AI combination led to performance losses, likely because humans were unable to integrate the suggestions provided by the AI.
Conversely, when the human outperformed the AI alone, there was some synergy and human–AI combination led to performance gains, likely because this time humans were better at integrating the AI suggestions.
A meta-analysis of 106 studies just published in Nature reports an interesting result:
On average, there was no synergy: Human–AI combinations did not perform better than both humans and AI.
In particular, when the AI alone outperformed the human alone, the human–AI combination led to performance losses, likely because humans were unable to integrate the suggestions provided by the AI.
Conversely, when the human outperformed the AI alone, there was some synergy and human–AI combination led to performance gains, likely because this time humans were better at integrating the AI suggestions.
Heuristics AI
28 Oct, 14:54
GDDR7 Memory Supercharges AI Inference https://semiengineering.com/gddr7-memory-supercharges-ai-inference/
Heuristics AI
28 Oct, 05:35
Why Explainability Matters in AI. Not because we’re curious. Because we… | by Uri Merhav | Oct, 2024
https://towardsdatascience.com/why-explainability-matters-in-ai-840144df418e
https://towardsdatascience.com/why-explainability-matters-in-ai-840144df418e
Heuristics AI
26 Oct, 06:21
https://github.com/huggingface/diffusers/tree/main/examples/research_projects/flux_lora_quantization
Heuristics AI
24 Oct, 13:40
AI Mathematical Olympiad - Progress Prize 2 Overview:
The goal of this competition is to create algorithms and models that can solve tricky math problems written in LaTeX format. Your participation will help to advance AI models’ mathematical reasoning skills and drive frontier knowledge. Competition Host
AI|MO
Prizes & Awards
$2,117,152
Awards Points & Medals link:https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/overview
The goal of this competition is to create algorithms and models that can solve tricky math problems written in LaTeX format. Your participation will help to advance AI models’ mathematical reasoning skills and drive frontier knowledge. Competition Host
AI|MO
Prizes & Awards
Awards Points & Medals link:https://www.kaggle.com/competitions/ai-mathematical-olympiad-progress-prize-2/overview
Heuristics AI
24 Oct, 07:48
https://youtu.be/xL6Y0dpXEwc?si=N57_sAxJDNszb-h7
How could machines reach human level intelligence? Talk by Yann lecun at Columbia University
How could machines reach human level intelligence? Talk by Yann lecun at Columbia University
Heuristics AI
24 Oct, 07:25
#tools
#MLSecOps
"Insights and Current Gaps in Open-Source LLM Vulnerability Scanners: A Comparative Analysis", 2024.
#MLSecOps
"Insights and Current Gaps in Open-Source LLM Vulnerability Scanners: A Comparative Analysis", 2024.
Heuristics AI
22 Oct, 17:19
NVidia just invented 15x faster Transformer: nGPT
Read the Paper 📜
Abstract: We propose a novel neural network architecture, the normalized Transformer
(nGPT) with representation learning on the hypersphere. In nGPT, all vectors
forming the embeddings, MLP, attention matrices and hidden states are unit norm
normalized. The input stream of tokens travels on the surface of a hypersphere,
with each layer contributing a displacement towards the target output predictions.
These displacements are defined by the MLP and attention blocks, whose vector
components also reside on the same hypersphere. Experiments show that nGPT
learns much faster, reducing the number of training steps required to achieve the
same accuracy by a factor of 4 to 20, depending on the sequence length.
Read the Paper 📜
Abstract: We propose a novel neural network architecture, the normalized Transformer
(nGPT) with representation learning on the hypersphere. In nGPT, all vectors
forming the embeddings, MLP, attention matrices and hidden states are unit norm
normalized. The input stream of tokens travels on the surface of a hypersphere,
with each layer contributing a displacement towards the target output predictions.
These displacements are defined by the MLP and attention blocks, whose vector
components also reside on the same hypersphere. Experiments show that nGPT
learns much faster, reducing the number of training steps required to achieve the
same accuracy by a factor of 4 to 20, depending on the sequence length.
Heuristics AI
18 Oct, 07:15
Guide to ML Engineer Job Hunting
https://www.yuan-meng.com/posts/mle_interviews/
https://www.yuan-meng.com/posts/mle_interviews/
Heuristics AI
17 Oct, 19:47
GitHub - KellerJordan/modded-nanogpt: NanoGPT (124M) quality in 2.67B tokens
https://github.com/KellerJordan/modded-nanogpt
https://github.com/KellerJordan/modded-nanogpt
Heuristics AI
15 Oct, 17:48
PyTorch Optimizers Aren’t Fast Enough. Try These Instead | by Benjamin Bodner | Oct, 2024 | Towards Data Science
https://towardsdatascience.com/pytorch-optimizers-arent-fast-enough-try-these-instead-61a1350e3eac
https://towardsdatascience.com/pytorch-optimizers-arent-fast-enough-try-these-instead-61a1350e3eac
Heuristics AI
14 Oct, 14:42
Folding Large Proteins by Ultra-Deep Learning | Proceedings of the 8th ACM International Conference on Bioinformatics, Computational Biology,and Health Informatics
https://dl.acm.org/doi/abs/10.1145/3107411.3107456
https://dl.acm.org/doi/abs/10.1145/3107411.3107456
Heuristics AI
14 Oct, 11:34
https://x.com/s_batzoglou/status/1845066153506267346?t=Z7sIrgoOhY8Zl4iqpGgQRw&s=19
Heuristics AI
12 Oct, 18:40
https://www.deeplearning.ai/short-courses/introducing-multimodal-llama-3-2/
#weeknd_snacks
#weeknd_snacks
Heuristics AI
11 Oct, 10:26
https://www.google.com/about/careers/applications/jobs/results/123161199633146566-ai-innovation-and-research-machine-learning-software-engineer
#Hiring
#Hiring
Heuristics AI
11 Oct, 10:21
Making Deep Learning Go Brrrr From First Principles
https://horace.io/brrr_intro.html
https://horace.io/brrr_intro.html
Heuristics AI
10 Oct, 08:46
https://www.youtube.com/playlist?list=PL1EviN0AzZsbebjAD7pkzdDhgvdeL_JSa&jct=eb_V3roCChyPzDsB_mu40A let us colaborate on this playlist. Add youtube videos on maths for ML.
2,160
subscribers
101
photos
27
videos
