techyrushabh Profile Banner
Techy Rushabh Profile
Techy Rushabh

@techyrushabh

Followers
2
Following
11
Media
4
Statuses
374

Coding, Roadmap, Solutions, Projects and more...

India
Joined February 2022
Don't wanna be here? Send us removal request.
@akku92
Akash Vatsa 
6 months
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
@techyrushabh
Techy Rushabh
6 months
0
0
0
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
Ever wonder why even *experts* mess up? 🤔 New paper uses expert failures to make LLM agents SMARTER! Exploring Expert Failures (EEF) tunes agents by learning from mistakes. #AI #LLM
1
0
0
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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
Tweet card summary image
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
Ever wondered if robots could REALLY hold your hand? 👋 Researchers built RUKA, a 3D-printed humanoid hand that's compact, strong, & learns to grasp like us using motion capture data! Open-source design & code available! #Robotics #AI
1
0
0
@techyrushabh
Techy Rushabh
8 months
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
Tweet card summary image
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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
Tweet card summary image
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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
Tweet card summary image
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
@techyrushabh
Techy Rushabh
8 months
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
@techyrushabh
Techy Rushabh
8 months
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