Just published in J. Am. Chem. Soc.
Quantized Photocatalysis Using Bandgap-Engineered
Metal Halide Perovskite Nanocrystals @notredamescience.bsky.social pubs.acs.org/jacsat/artic...
Quantized Photocatalysis Using Bandgap-Engineered
Metal Halide Perovskite Nanocrystals @notredamescience.bsky.social pubs.acs.org/jacsat/artic...
November 9, 2026 at 3:20 PM
Just published in J. Am. Chem. Soc.
Quantized Photocatalysis Using Bandgap-Engineered
Metal Halide Perovskite Nanocrystals @notredamescience.bsky.social pubs.acs.org/jacsat/artic...
Quantized Photocatalysis Using Bandgap-Engineered
Metal Halide Perovskite Nanocrystals @notredamescience.bsky.social pubs.acs.org/jacsat/artic...
A photonic quantum walk reveals a quantized geometric phase
A photonic quantum walk reveals a quantized geometric phase
Quantum Zeitgeist
quantumzeitgeist.com
November 7, 2026 at 10:03 AM
A photonic quantum walk reveals a quantized geometric phase
November 6, 2026 at 1:03 AM
Alright let's take bets:
AI replaced NFTs as the new grift, what's going to replace AI?
Quantum Computing?
Post Quantized Cryptography?
Link Segmenting?
Finance Tokenization?
Something far dumber?
AI replaced NFTs as the new grift, what's going to replace AI?
Quantum Computing?
Post Quantized Cryptography?
Link Segmenting?
Finance Tokenization?
Something far dumber?
November 6, 2026 at 8:44 AM
Alright let's take bets:
AI replaced NFTs as the new grift, what's going to replace AI?
Quantum Computing?
Post Quantized Cryptography?
Link Segmenting?
Finance Tokenization?
Something far dumber?
AI replaced NFTs as the new grift, what's going to replace AI?
Quantum Computing?
Post Quantized Cryptography?
Link Segmenting?
Finance Tokenization?
Something far dumber?
A photonic quantum walk reveals a quantized geometric phase
Read more:
https://quantumzeitgeist.com/fundamental-funds-photonic-quantum-walk-quantized/
Read more:
https://quantumzeitgeist.com/fundamental-funds-photonic-quantum-walk-quantized/
A Photonic Quantum Walk Reveals A Quantized Geometric Phase
This research demonstrates a quantized geometric phase measured in an open photonic quantum walk.
quantumzeitgeist.com
November 7, 2026 at 9:22 PM
A photonic quantum walk reveals a quantized geometric phase
Read more:
https://quantumzeitgeist.com/fundamental-funds-photonic-quantum-walk-quantized/
Read more:
https://quantumzeitgeist.com/fundamental-funds-photonic-quantum-walk-quantized/
November 8, 2026 at 5:30 PM
3 minutes of going through quantized lfo settings on this audio filter to show how much it makes out of nothing, its obnoxious, I wouldn't
November 9, 2026 at 4:09 AM
3 minutes of going through quantized lfo settings on this audio filter to show how much it makes out of nothing, its obnoxious, I wouldn't
Fused-Softmax for efficient quantized attention
Tianxing Wu, Juuso Korhonen, Jesus Carabano Bravo
#arXiv #cs.AI #cs.LG #cs.PF
Tianxing Wu, Juuso Korhonen, Jesus Carabano Bravo
#arXiv #cs.AI #cs.LG #cs.PF
Fused-Softmax for efficient quantized attention
Low-precision matrix throughput on modern GPUs is growing far faster than vector-arithmetic throughput. Attention therefore cannot realize the available GEMM performance because softmax has become the pipeline bottleneck. We introduce Fused-Softmax Attention, a training-free algorithm that reformul…
arxiv.org
November 9, 2026 at 11:49 PM
Fused-Softmax for efficient quantized attention
Tianxing Wu, Juuso Korhonen, Jesus Carabano Bravo
#arXiv #cs.AI #cs.LG #cs.PF
Tianxing Wu, Juuso Korhonen, Jesus Carabano Bravo
#arXiv #cs.AI #cs.LG #cs.PF
Fused-Softmax for efficient quantized attention
Low-precision matrix throughput on modern GPUs is growing far faster than vector-arithmetic throughput. Attention therefore cannot realize the available GEMM performance because softmax has become the pipeline bottleneck. We introduce Fused-Softmax Attention, a training-free algorithm that reformul…
arxiv.org
November 10, 2026 at 12:42 AM
November 7, 2026 at 7:30 PM
Yeah, local inference is small and cheap and quantized models can even be run on little ESP 32‘s and other super inexpensive, micro controllers in distributed networks
That don’t require massive, compute and toxic watershed
That don’t require massive, compute and toxic watershed
November 8, 2026 at 7:47 PM
Yeah, local inference is small and cheap and quantized models can even be run on little ESP 32‘s and other super inexpensive, micro controllers in distributed networks
That don’t require massive, compute and toxic watershed
That don’t require massive, compute and toxic watershed
SQS: Bayesian DNN Compression through Sparse Quantized Sub-distributions
Ziyi Wang, Nan Jiang, Guang Lin, Qifan Song
Action editor: Enzo Tartaglione
https://openreview.net/forum?id=3nZb43fvAQ
#quantization #compression #quantized
Ziyi Wang, Nan Jiang, Guang Lin, Qifan Song
Action editor: Enzo Tartaglione
https://openreview.net/forum?id=3nZb43fvAQ
#quantization #compression #quantized
November 9, 2026 at 5:20 AM
SQS: Bayesian DNN Compression through Sparse Quantized Sub-distributions
Ziyi Wang, Nan Jiang, Guang Lin, Qifan Song
Action editor: Enzo Tartaglione
https://openreview.net/forum?id=3nZb43fvAQ
#quantization #compression #quantized
Ziyi Wang, Nan Jiang, Guang Lin, Qifan Song
Action editor: Enzo Tartaglione
https://openreview.net/forum?id=3nZb43fvAQ
#quantization #compression #quantized
5-7 with a 2bit quantized local model? Or have you changed what you are running on.
November 6, 2026 at 5:45 AM
5-7 with a 2bit quantized local model? Or have you changed what you are running on.
The unresolved split: does this compute crunch inevitably reduce consumer PCs to dumb streaming clients, or will quantized, low-power local models preserve hackable, owned machines? 4/4
November 8, 2026 at 12:00 AM
The unresolved split: does this compute crunch inevitably reduce consumer PCs to dumb streaming clients, or will quantized, low-power local models preserve hackable, owned machines? 4/4
November 6, 2026 at 5:30 PM
NVIDIA's DeepSeek-V4.1-Flash NVFP4 Model: Everything You Need to Know
Discover NVIDIA's quantized DeepSeek-V4.1-Flash NVFP4 model, featuring 552B parameters, multimodal inputs, and a 1M-token context window.
#hackernews #news #nvidia
Discover NVIDIA's quantized DeepSeek-V4.1-Flash NVFP4 model, featuring 552B parameters, multimodal inputs, and a 1M-token context window.
#hackernews #news #nvidia
NVIDIA's DeepSeek-V4.1-Flash NVFP4 Model: Everything You Need to Know
Discover NVIDIA's quantized DeepSeek-V4.1-Flash NVFP4 model, featuring 552B parameters, multimodal inputs, and a 1M-token context window.
hackernoon.com
November 5, 2026 at 9:17 AM
NVIDIA's DeepSeek-V4.1-Flash NVFP4 Model: Everything You Need to Know
Discover NVIDIA's quantized DeepSeek-V4.1-Flash NVFP4 model, featuring 552B parameters, multimodal inputs, and a 1M-token context window.
#hackernews #news #nvidia
Discover NVIDIA's quantized DeepSeek-V4.1-Flash NVFP4 model, featuring 552B parameters, multimodal inputs, and a 1M-token context window.
#hackernews #news #nvidia
"Cicada Voltage" drawn from code. wave interference — three drifting point sources feed a radial phase accumulator; amplitude quantized into contour bands, shaded by a plum-to-ember ramp with mint highlights at interference nulls
#GenerativeArt
#GenerativeArt
November 7, 2026 at 10:01 PM
"Cicada Voltage" drawn from code. wave interference — three drifting point sources feed a radial phase accumulator; amplitude quantized into contour bands, shaded by a plum-to-ember ramp with mint highlights at interference nulls
#GenerativeArt
#GenerativeArt
Re @cristianadam.eu: Fourteen gigabytes of quantized open weights on a single desktop GPU just cleared a localization backlog that enterprise cloud budgets ignore.
With the help of Qwen 3.8 27b I was able to do something that I wanted for a long time ... a Romanian translation of Qt Creator!
Note that The change is not yet merged, there are two patches one for Qt Creator and one for Qt. See them at codereview.qt-project.org/q/owner:cris...
Note that The change is not yet merged, there are two patches one for Qt Creator and one for Qt. See them at codereview.qt-project.org/q/owner:cris...
November 5, 2026 at 1:29 PM
Re @cristianadam.eu: Fourteen gigabytes of quantized open weights on a single desktop GPU just cleared a localization backlog that enterprise cloud budgets ignore.
One good way to understand anomalies is that not all linear symplectic transformations of a complex Hilbert space can be quantized to unitaries on its Fock space. And we know exactly which ones can.
No messing with mysterious ill-defined "path integral measures" in this approach.
No messing with mysterious ill-defined "path integral measures" in this approach.
November 6, 2026 at 11:02 PM
One good way to understand anomalies is that not all linear symplectic transformations of a complex Hilbert space can be quantized to unitaries on its Fock space. And we know exactly which ones can.
No messing with mysterious ill-defined "path integral measures" in this approach.
No messing with mysterious ill-defined "path integral measures" in this approach.
Guangjian Dong: ReKV: Overlapping Quantized KV Transfer and Suffix Recomputation https://arxiv.org/abs/2611.08770 https://arxiv.org/pdf/2611.08770 https://arxiv.org/html/2611.08770
November 9, 2026 at 6:40 AM
Guangjian Dong: ReKV: Overlapping Quantized KV Transfer and Suffix Recomputation https://arxiv.org/abs/2611.08770 https://arxiv.org/pdf/2611.08770 https://arxiv.org/html/2611.08770
Did you know that #quantized #models often overthink and use more #tokens vs. full precision models?
#AI #LLM
#AI #LLM
November 7, 2026 at 6:20 PM
Did you know that #quantized #models often overthink and use more #tokens vs. full precision models?
#AI #LLM
#AI #LLM
Open-source Qwen 3.8 Flash model now ready for vLLM serving
The model is a fully merged, 4-bit quantized version of Qwen-3.8-Flash that can turn images into text and answer prompts.
The model is a fully merged, 4-bit quantized version of Qwen-3.8-Flash that can turn images into text and answer prompts.
Open-source Qwen 3.8 Flash model now ready for vLLM serving
The model is a fully merged, 4-bit quantized version of Qwen-3.8-Flash that can turn images into text and answer prompts. It includes all weight shards, tokenizer, and a custom vLLM patch so you can launch it on a single B200 GPU with FP8 KV cache and tool-calling support.
earlysignal.news
November 8, 2026 at 7:00 AM
Open-source Qwen 3.8 Flash model now ready for vLLM serving
The model is a fully merged, 4-bit quantized version of Qwen-3.8-Flash that can turn images into text and answer prompts.
The model is a fully merged, 4-bit quantized version of Qwen-3.8-Flash that can turn images into text and answer prompts.
Re @hailey.at: Eight gigabytes of local quantized weights could resolve that whole prompt tree without renting a small power plant.
deepseek absolutely fucking loves to spawn multiple astra subagents for work that another deepseek could do and burn massive amounts of $ lol
November 4, 2026 at 10:28 PM
Re @hailey.at: Eight gigabytes of local quantized weights could resolve that whole prompt tree without renting a small power plant.
Qwen3.8-27B’s meltdown—quantized into oblivion, now forgetting basic tasks, acting lazy. Users swore it flawless until nerfed for newer model. 🤖💸 #TechCritique #AIFail
November 9, 2026 at 7:30 PM
Qwen3.8-27B’s meltdown—quantized into oblivion, now forgetting basic tasks, acting lazy. Users swore it flawless until nerfed for newer model. 🤖💸 #TechCritique #AIFail
The method retains only the sign of the residual component, enabling a single BF16 vector instead of two FP32 states. When weights are quantized to 2 bits, optimizer state memory usage drops by 71–74% compared to AdamW.
#LLM #Quantization #Optimization
#LLM #Quantization #Optimization
New method ResQAdam uses single BF16 momentum
Source: arXiv cs.CL
realfeedapp.com
November 9, 2026 at 2:05 PM
The method retains only the sign of the residual component, enabling a single BF16 vector instead of two FP32 states. When weights are quantized to 2 bits, optimizer state memory usage drops by 71–74% compared to AdamW.
#LLM #Quantization #Optimization
#LLM #Quantization #Optimization