Top Tweets for #InferenceOptimization
Minimax-h3-image-pdmd-2step, a 33B model realtime from akhaliq
pDMD 2-Step LoRA for MiniMaxAI/MiniMax-H3-Image, a 33.1B-parameter diffusion transformer, on...
Read more → https://t.co/YyjKJ0kS2N…
#AI #ComputerVision #DeepLearning #InferenceOptimization

📰 EngineAgent Automates On-Device LLM Enablement with 75% Success Rate
EngineAgent achieves 75% end-to-end success enabling new LLMs on mobile hardware, compared with a 33.3% baseline...
Read more → https://t.co/OBmBGTwQ2g
#AI #AIAgents #AiHardware #InferenceOptimization

📰 MVRD distillation outperforms fine-tuning for VLM spatial reasoning
MVRD distills cross-view cosine similarities into VLMs, beating supervised fine-tuning on spatial reasoning with...
Read more → https://t.co/BtoCv7MG9w
#AI #AiSafety #ComputerVision #InferenceOptimization

📰 Tether AI Research BotPsy 0.8B beats 2B models on agentic benchmarks
Tether AI Research’s 0.8B model posts 45.0 on six agentic benchmarks, above Qwen3.5-2B-Think at 38.8, enabling...
Read more → https://t.co/CvHeWIacKx
#AI #AIAgents #InferenceOptimization #OpenSourceAI

📰 ModernBERT pooling-aware training lifts nDCG at 75% fewer vectors
Tom Aarsen published a technical post on Hugging Face on November 11, 2026, introducing token pooling-aware training for...
Read more → https://t.co/ZN9JbX2HDL
#AI #FineTuning #InferenceOptimization #LearnAI

📰 Hugging Face Hub Adds Google Cloud Marketplace Billing
Hugging Face orgs can now route Hub and compute billing through Google Cloud Marketplace, consolidating payments onto an...
Read more → https://t.co/ctOLqrX8hA
#AiHardware #AIIndustry #AITools #InferenceOptimization

📰 SIRIS: 13 LLMs Identify Simulator Flaws but Still Abandon Valid Picks
Evaluating 13 LLMs on serving design, SIRIS finds that detecting a simulator flaw does not ensure a correct choice, and...
Read more → https://t.co/T7nrbkAYa4
#AI #AIAgents #InferenceOptimization #LLM

📰 Decision Models in Hugging Face
The decision model interface includes Choice, Score, and Noul primitives, enabling structured evaluations from a state without...
Read more → https://t.co/VQm4TfXXq6
#AI #AITools #InferenceOptimization #LLM

📰 Flide runs decision inference 9.5x faster than vLLM and SGLang
Flide uses ForkAttention to branch decision model queries in one stream, running up to 9.5x faster than vLLM and SGLang...
Read more → https://t.co/QZFbO3ALtq
#AI #AIAgents #AiHardware #InferenceOptimization

📰 Google DeepMind Announces Gemini Robotics 2.5 for Physical AI
Google DeepMind introduced Gemini Robotics 2.5, a physical AI model that learns hundreds of everyday skills like folding laundry...
Read more → https://t.co/y0UqH7SnaK
#AI #AIAgents #InferenceOptimization #LLM

📰 Seeing What Matters: Benchmarking Task-Conditioned Actionable Spatial Belief for Manipulation
The reference pipeline hits 98.8 percent closed-loop success while seven VLMs score...
Read more → https://t.co/e7IdIcnspl
#AIIndustry #InferenceOptimization #AIReasoning #Robotics

📰 DEQ-Stereo Brings Adaptive Convergence to Video Stereo Matching
DEQ-Stereo uses adaptive convergence and temporal state transfer to claim state-of-the-art video stereo...
Read more → https://t.co/fXGx5kNlFg
#AIIndustry #ComputerVision #DeepLearning #InferenceOptimization

📰 Scratchpad Patching Enables 16x KV-Cache Cuts in Byte-Level LLMs
Transient scratchpads inside byte patches let tokenizer-free models match baseline quality with 16x smaller KV-cache and 3-4x...
Read more → https://t.co/34w1CZEKGe
#AI #AiHardware #InferenceOptimization #LLM

📰 Amazon HOPE Prunes MoE Experts Using Second-Order Interactions
Amazon’s HOPE uses a quadratic program over expert pairs, beating first-order pruning by up to 6.1 percent on agentic coding at...
Read more → https://t.co/cWgANizett
#AI #AiHardware #InferenceOptimization #LLM

📰 Uni-LaSpec Accelerates VLM Inference Up to 5.11x via Speculative Draft
Uni-LaSpec selects eight visual tokens via cross-layer voting to draft, delivering up to 5.11x text throughput while...
Read more → https://t.co/fkoPYFj6Z9
#AI #AiHardware #InferenceOptimization #LLM

📰 Sam Power Surveys Poincare and Log-Sobolev Inequalities in Langevin
Sam Power’s tutorial anchors Poincare and log-Sobolev inequalities in Langevin diffusion, linking them to coupling and...
Read more → https://t.co/HjcUA7E0Bk
#AI #DataScience #InferenceOptimization #LearnAI

📰 A Survey on Continuous Diffusion and Flow Language Models
Sophia Tang, Chanhyuk Lee, Floor Eijkelboom, and Jinwoo Kim published a 37-page survey on continuous diffusion and flow language...
Read more → https://t.co/PCl1s0UtZa
#AI #DeepLearning #InferenceOptimization #LLM

📰 Muir linear attention extrapolates 2,560x to 10M-token streams
Muir is a linear attention model that, at 1.34B parameters, lowers perplexity relative to Gated DeltaNet-2 and Transformer...
Read more → https://t.co/kG3c3geQ8X
#AI #DeepLearning #InferenceOptimization #LLM

📰 Failure Prediction and Classification in Vision-Language-Action Models Via Calibrated Latent Probing
The authors present VLA-Tug, a lightweight probe over the latent layers of a...
Read more → https://t.co/3EN2FohUgA
#AI #AiSafety #InferenceOptimization #Multimodal

📰 Hugging Face livestream covers llama.cpp and vLLM quantization
Hugging Face hosts a five-speaker livestream on quantizing models with llama.cpp and vLLM to shrink models to fit in a...
Read more → https://t.co/XOvZeeeJfh
#AI #InferenceOptimization #LearnAI #OpenSourceAI

Last Seen Hashtags on Sotwe
brosis
Seen from Kenya
parvathibabu
Seen from India
ตาก
Seen from Thailand
anxiouspanda
varsello
Seen from Italy
شركة_الحمراء_لخدمات_حجاج_الداخل
Seen from United States
yenge(**************)
Seen from Turkey
monkeyapp
Seen from Netherlands
public handjob
Seen from Turkey
premiosecovidrio
Seen from United States
Trends for you
Most Popular Users

Elon Musk 
@elonmusk
241.3M followers

Barack Obama 
@barackobama
118.8M followers

Cristiano Ronaldo 
@cristiano
114.5M followers

Donald J. Trump 
@realdonaldtrump
111.9M followers

Narendra Modi 
@narendramodi
107M followers

Rihanna 
@rihanna
98.7M followers

NASA 
@nasa
92.3M followers

Justin Bieber 
@justinbieber
91.8M followers

KATY PERRY 
@katyperry
90M followers

Taylor Swift 
@taylorswift13
84M followers

Lady Gaga 
@ladygaga
75.5M followers

Virat Kohli 
@imvkohli
73.4M followers

Kim Kardashian 
@kimkardashian
70.8M followers

YouTube 
@youtube
68.7M followers

Neymar Jr 
@neymarjr
66.4M followers

Bill Gates 
@billgates
65.2M followers

Selena Gomez 
@selenagomez
63.1M followers

The Ellen Show
@theellenshow
62.2M followers

CNN 
@cnn
61.7M followers

X 
@x
60.6M followers
