Techy Rushabh
@techyrushabh
Followers
2
Following
11
Media
4
Statuses
374
Coding, Roadmap, Solutions, Projects and more...
India
Joined February 2022
I was in the same damn flight 2 hours before it took off from AMD. I came in this from DEL-AMD. Noticed unusual things in the place.Made a video to tweet to @airindia i would want to give more details. Please contact me. @flyingbeast320 @aajtak @ndtv @Boeing_In #planecrash #AI171
5K
26K
114K
https://t.co/IB83k0NSK1 This insight might help to investigate #AirIndiaCrash #AirIndiaFlightCrash
@airindia
0
0
0
Want to make AI image editing *actually* follow complex instructions? This new paper, "Complex-Edit," introduces a benchmark pushing the limits! Think "turn my cat into a cyberpunk detective" level stuff. 🤯 #AI #ImageEditing
0
0
0
Mind-blowing! 🤯 This "EEF" method achieves state-of-the-art results in WebShop & SciWorld. Turns out, understanding *how* experts fail is key to better #MachineLearning agents! 🚀 https://t.co/ZAUcC3wVNr
0
0
0
Antidistillation Sampling lets you make a model *less* distillable without hurting its performance! That's right - keep your edge AND protect your IP. Is this the future of #MachineLearning security? 🤔 https://t.co/PItp3jqrCk
0
0
0
Think your model is safe from being copied via distillation? Think again! 🤯 "Antidistillation Sampling" introduces a sneaky way to poison those reasoning traces & protect your secret sauce. Check it out! #AI #ModelSecurity
2
0
0
Unlocking the black box! This research successfully created a "readable twin" for an image recognition model. Imagine applying this to high-stakes decisions - making AI more transparent & trustworthy! Huge implications for responsible #MachineLe... https://t.co/r8A9YxBLH2
arxiv.org
Creating responsible artificial intelligence (AI) systems is an important issue in contemporary research and development of works on AI. One of the characteristics of responsible AI systems is...
0
0
0
Ever wonder what your AI is *actually* thinking? 🤔 New paper introduces "readable twins" for deep learning models! Like digital twins, but for AI - turning complex models into understandable flowcharts. Check it out! #AI #ExplainableAI
1
0
0
MIB reveals standardized ways to compare interpretability methods, meaning we can FINALLY track real progress in understanding AI. A huge step toward safer & more reliable models! 🚀 #ML #AIexplainability
https://t.co/F8FQeDHuP5
0
0
0
Ever wonder how neural nets *actually* think? 🤔 New paper "MIB" offers a benchmark to test if we can truly understand their inner workings & find causal pathways! Ready to peek under the hood? 🧰 #AI #MechanisticInterpretability
1
0
0
RUKA isn't just a pretty hand! 💪 It boasts superior reach, durability & strength. Learning-based control opens doors for more dexterous & adaptable robots in the real world. Imagine the possibilities! #MachineLearning #Robotics
https://t.co/iAdgmLipJn
0
0
0
Sleep-time compute can drastically cut test-time needs & improve accuracy, especially in reasoning tasks! The better the query predictability, the bigger the payoff. Agentic SWE case study included! 🤩 #MachineLearning #AIResearch
https://t.co/n4uWPlZQ4S
arxiv.org
Scaling test-time compute has emerged as a key ingredient for enabling large language models (LLMs) to solve difficult problems, but comes with high latency and inference cost. We introduce...
0
0
1
Ever wish your AI could anticipate your questions and have the answer ready? 😴 This paper introduces "sleep-time compute," letting models prep offline to slash test-time costs & boost accuracy!🤯 #AI #EfficientAI
1
0
0
This paper introduces Miras, a framework that rethinks deep learning architectures as associative memory! Implications? Better language models, commonsense reasoning, & recall. Are we unlocking true understanding? Read more! [link to paper] #Mac... https://t.co/pcwXSSSAfT
arxiv.org
Designing efficient and effective architectural backbones has been in the core of research efforts to enhance the capability of foundation models. Inspired by the human cognitive phenomenon of...
0
0
1
Ever wonder if your neural network is just *memorizing* instead of *learning*? 🤔 This new paper, "It's All Connected," dives deep into test-time memorization using associative memory & attention. Get ready to rethink how your models "remember"! #AI #DeepLearning
1
0
1
No more black box! PerceptionLM's study found gaps in current video training data & releases 2.8M NEW human-labeled video Q&A/captions! HUGE step towards transparent & accurate visual AI. Ready to dive in? 🚀 #MachineLearning #OpenSourceAI
https://t.co/o5JwHXrQCp
arxiv.org
Vision-language models are integral to computer vision research, yet many high-performing models remain closed-source, obscuring their data, design and training recipe. The research community has...
0
0
0
Ever wished AI could REALLY "see" like we do? 👀 New paper "PerceptionLM" tackles that! They're releasing open-source data & models to boost detailed image/video understanding. Think less vague, more nuanced AI vision! #AI #ComputerVision
1
0
0
Ever wonder if AI can *truly* understand space? This paper explores how neural networks learn spatial concepts just by navigating open environments! 🤯 Embodied World Models Emerge from Navigational Task in Open-Ended Environments by Jin & Jia. #AI #EmbodiedAI
0
0
0