Finally a benchmark I have been working on. Fable wins unsurprisingly.
Its a benchmark for reverse engineering byte code of old systems (like N64) to a source code that is *byte identical* when compiled.
So it is very hard to do but very easy to measure. And for succeeding you need to have a deep understanding of the systems and software and most importantly you need to be able to make the best of available tools. Details in the page.
There are initial results. Needs reliability improvements and averaging some runs.
Details in the page: https://t.co/uS1uIYMsb5
#aibenchmark #retrobench #fable5 #GPT56So #benchmark
@karpathy I've been thinking the same. Off loading the workload to CLIs are much more efficient and effective in terms of context use. It is also predictable on what agent can do. I've built a skill manager in that perspective.
https://t.co/4vgV8DIhSb
We've all been playing around with agent skills for a while. Lately, I've been thinking about how to manage skills, use context efficiently and have agents do the skill work.I've proceeded building Skern, a system-wide skill registry for AI coding agents.
https://t.co/aT4XlU54rs
What makes it different:
- Portable SKILL.md format, no proprietary config
- Fuzzy overlap detection to prevent duplicates
- --json on every command so agents can drive the CLI themselves
- Tool-forming loop: agents scaffold + reuse skills automatically
@zhu_zhaocheng@iclr_conf If the paper is accepted somehow, I think the paper can be removed and it's doi pointer shall be an empty page or a page describing the plagiarism. https://t.co/Xj2aO7PfWA
@mrdbourke Folks already gave loads of resources and explanations about why linear functions are inferior. Specifically for ReLU, I'd recommend you to read about human neural activations/pulses, and maybe experiments of Hubel and Wiesel.
@OguzhanUgur@oguzhanugur trt YouTubeda belli oranda restore edilmiş halini paylaşmıştı (sanırım geçen senelerde) ama restorasyonun ayarı kaçmış, bazı bölümler çıkarılmıştı, örneğin Albay Reşat Bey'in Gazi Paşa ile telefonla konuştuğu ve intihar ettiği sahne kesilmiş olarak yayınlandı.
Presenting TiDE, a time-series dense encoder for long-term time-series forecasting that enjoys the simplicity and speed of linear models while also being able to handle covariates and non-linear dependencies. Learn more →https://t.co/9y8kmN2lV2
We have open sourced the #biomedical#signalprocessing library we have been using as a team with slight improvements. Recommended especially if you are working with ECG, PPG, EDA or ACC signals. Comparison with other libraries in the pic #opensource https://t.co/xUqJtB1aTu (1)
// Stable Diffusion, Explained //
You've seen the Stable Diffusion AI art all over Twitter.
But how does Stable Diffusion _work_?
A thread explaining diffusion models, latent space representations, and context injection:
1/15
A simple Python package for handling bounding box format conversions: pybboxes
Currently supports: albumentations, coco, fiftyone, voc and yolo formats 🚀
Github: https://t.co/bVxh7MdhQH
Jury is out for Natural Language Generation!
As @obsscomtr Machine Learning team we are open sourcing Jury, a toolkit to easily evaluate your #nlp models that generates text.
Github: https://t.co/dyKCCJ20yj
Blog: https://t.co/yubboOfMuo
Kudos to @adecavusoglu for heavy lifting.