Oh, really? Get TF in line, bro, there's a big group of people seeking "regularization", whatever TF that is.
November 15, 2026 at 6:42 PM
Oh, really? Get TF in line, bro, there's a big group of people seeking "regularization", whatever TF that is.
yeah, codependent overfitting. this is why ceos should be subject to dropout regularization and bagging
November 13, 2026 at 4:11 PM
yeah, codependent overfitting. this is why ceos should be subject to dropout regularization and bagging
Flávio Bolsonaro: I will remain here in the US. I will regularize my immigration status.
Once I complete this legal regularization process, I do not see how any government, regardless of who takes office after 2028, could simply reverse it out of revenge, resentment or hatred
x.com/RpsAgainstTr...
Once I complete this legal regularization process, I do not see how any government, regardless of who takes office after 2028, could simply reverse it out of revenge, resentment or hatred
x.com/RpsAgainstTr...
Republicans against Trump (@RpsAgainstTrump) on X
Flávio Bolsonaro: I will remain here in the US. I will regularize my immigration status.
Once I complete this legal regularization process, I do not see how any government, regardless of who takes of...
x.com
November 15, 2026 at 6:35 PM
Flávio Bolsonaro: I will remain here in the US. I will regularize my immigration status.
Once I complete this legal regularization process, I do not see how any government, regardless of who takes office after 2028, could simply reverse it out of revenge, resentment or hatred
x.com/RpsAgainstTr...
Once I complete this legal regularization process, I do not see how any government, regardless of who takes office after 2028, could simply reverse it out of revenge, resentment or hatred
x.com/RpsAgainstTr...
When Covariance Is Not Enough: Kernel-Based Regularization for Single-Cell Counterfactuals #SingleCell 🧪🧬🖥️
https://arxiv.org/abs/2611.13748
https://arxiv.org/abs/2611.13748
November 11, 2026 at 5:01 AM
When Covariance Is Not Enough: Kernel-Based Regularization for Single-Cell Counterfactuals #SingleCell 🧪🧬🖥️
https://arxiv.org/abs/2611.13748
https://arxiv.org/abs/2611.13748
The blog shows the method, training code, checkpoints & evaluation. I also build on Stefan Josef's Learn to Pool and @nohtow.bsky.social @ @lightonai.bsky.social's pooling regularization work.
I also use Late Interaction Kernels by Tony Wu and Aurélien Lac @ H Company. Big kudos to all!
I also use Late Interaction Kernels by Tony Wu and Aurélien Lac @ H Company. Big kudos to all!
November 11, 2026 at 3:05 PM
The blog shows the method, training code, checkpoints & evaluation. I also build on Stefan Josef's Learn to Pool and @nohtow.bsky.social @ @lightonai.bsky.social's pooling regularization work.
I also use Late Interaction Kernels by Tony Wu and Aurélien Lac @ H Company. Big kudos to all!
I also use Late Interaction Kernels by Tony Wu and Aurélien Lac @ H Company. Big kudos to all!
Struggle for Land Regularization at the Federal University of Ceará, Brazil
Struggle for Land Regularization at the Federal University of Ceará, Brazil
www.anarchistfederation.net
November 8, 2026 at 5:59 PM
Struggle for Land Regularization at the Federal University of Ceará, Brazil
Novel variational framework improves conditioning of Krylov subspaces in quantum algorithms by penalizing linear dependencies while maintaining Lanczos-like convergence—addressing a key numerical stability challenge in near-term quantum eigensolvers.
#QuantumAlgorithms #QuantumComputing #Research
#QuantumAlgorithms #QuantumComputing #Research
Variational Regularization of Krylov Subspaces for Quantum Eigensolvers
iq.fp2.dev
November 13, 2026 at 11:50 PM
Novel variational framework improves conditioning of Krylov subspaces in quantum algorithms by penalizing linear dependencies while maintaining Lanczos-like convergence—addressing a key numerical stability challenge in near-term quantum eigensolvers.
#QuantumAlgorithms #QuantumComputing #Research
#QuantumAlgorithms #QuantumComputing #Research
Researchers developed CARE, a regularization framework that uses conditioning signals to improve diffusion models. The method achieves 3.5x faster training on ImageNet with 19% lower FID. It works as a plug and play addition to existing approaches.
CARE: Condition-Aware Representation Regularization for Diffusion Models
Recent advances in diffusion models highlight the importance of representation regularization for improving sample quality and training efficiency. However, commonly used regularization methods often overlook the built-in conditions (such as labels or texts) which directly determine the generation target. In this work, we demonstrate how conditioning signals affect the feature distribution and introduce the CARE (Condition-Aware REpresentation regularization). CARE is a lightweight plug-and-play regularization framework that dynamically modulates feature distribution based on condition similarity. CARE leverages built-in conditioning signals to judiciously guide the representation space, promoting tighter feature clusters for similar conditions without relying on explicit alignment losses or external supervision. Empirically, CARE consistently improves both visual fidelity and convergence stability across both class-to-image and text-to-image tasks. On ImageNet, CARE achieves a 19.08\% reduction in FID in 400k training steps, leading to a 3.5$\times$ speed-up. When applied to text-to-image generation, CARE lowers FID by 16.61\% in 200k iterations and improves semantic alignment between generated samples and text prompts. Moreover, CARE can be seamlessly integrated with existing regularization methods, yielding additional performance gains.
arxiv.org
November 12, 2026 at 9:48 PM
Researchers developed CARE, a regularization framework that uses conditioning signals to improve diffusion models. The method achieves 3.5x faster training on ImageNet with 19% lower FID. It works as a plug and play addition to existing approaches.
Trump buddy Flávio Bolsonaro bragging that he will be safe here once he regularizes his immigration status;
"Once I complete this legal regularization process, I do not see how any government, regardless of who takes office after 2028, could reverse it"
* unless Brazil wants him extradited there
"Once I complete this legal regularization process, I do not see how any government, regardless of who takes office after 2028, could reverse it"
* unless Brazil wants him extradited there
November 15, 2026 at 4:33 PM
Trump buddy Flávio Bolsonaro bragging that he will be safe here once he regularizes his immigration status;
"Once I complete this legal regularization process, I do not see how any government, regardless of who takes office after 2028, could reverse it"
* unless Brazil wants him extradited there
"Once I complete this legal regularization process, I do not see how any government, regardless of who takes office after 2028, could reverse it"
* unless Brazil wants him extradited there
Boosting Robot Policy Robustness via Inner Product Regularization in Imitation Learning + Video
Introduction: Imitation learning often suffers from severe performance degradation under distribution shifts and covariate drift because policies frequently latch onto task-irrelevant state dimensions.…
Introduction: Imitation learning often suffers from severe performance degradation under distribution shifts and covariate drift because policies frequently latch onto task-irrelevant state dimensions.…
Boosting Robot Policy Robustness via Inner Product Regularization in Imitation Learning + Video
Introduction: Imitation learning often suffers from severe performance degradation under distribution shifts and covariate drift because policies frequently latch onto task-irrelevant state dimensions. Inner Product Regularization (InneR) addresses this limitation by shaping the state-space geometry through the policy Jacobian, enforcing locally consistent perturbation responses across neighboring states. By aligning action-space metrics with state-space perturbations without sacrificing expressiveness, this technique drastically improves real-world robot manipulation success rates.
undercodetesting.com
November 15, 2026 at 11:06 AM
Boosting Robot Policy Robustness via Inner Product Regularization in Imitation Learning + Video
Introduction: Imitation learning often suffers from severe performance degradation under distribution shifts and covariate drift because policies frequently latch onto task-irrelevant state dimensions.…
Introduction: Imitation learning often suffers from severe performance degradation under distribution shifts and covariate drift because policies frequently latch onto task-irrelevant state dimensions.…
regularization. That is an example of what I’m advocating for over “every spike is sacred.” And I want to clarify I’m not accusing you of that personally, I’m referring to a broad trend in the field I think is problematic
November 11, 2026 at 1:17 PM
regularization. That is an example of what I’m advocating for over “every spike is sacred.” And I want to clarify I’m not accusing you of that personally, I’m referring to a broad trend in the field I think is problematic
Broad regularization
November 8, 2026 at 9:00 PM
Broad regularization
Migrant Regularization as Climate Adaptation: Marc Riudavets-Barcons
NEP/RePEc link
to paper
d.repec.org
November 10, 2026 at 2:59 AM
Migrant Regularization as Climate Adaptation: Marc Riudavets-Barcons
🇺🇸🇧🇷 Flávio Bolsonaro on his plans after his contentious election loss: "I will remain here in the United States. I will regularize my immigration status. Once I complete this legal regularization process, I do not see how any government, regardless of who takes office
November 15, 2026 at 12:35 PM
🇺🇸🇧🇷 Flávio Bolsonaro on his plans after his contentious election loss: "I will remain here in the United States. I will regularize my immigration status. Once I complete this legal regularization process, I do not see how any government, regardless of who takes office
Alexandre Perrin, J\'er\'emie Messud, S\'ebastien Burdin, Matthieu Sauban\`ere
Exploring Variational Regularization of Krylov Subspaces
https://arxiv.org/abs/2611.09229
Exploring Variational Regularization of Krylov Subspaces
https://arxiv.org/abs/2611.09229
November 10, 2026 at 10:01 PM
Alexandre Perrin, J\'er\'emie Messud, S\'ebastien Burdin, Matthieu Sauban\`ere
Exploring Variational Regularization of Krylov Subspaces
https://arxiv.org/abs/2611.09229
Exploring Variational Regularization of Krylov Subspaces
https://arxiv.org/abs/2611.09229
Jiahao Qin, Yiwen Wang: M\"obius Interaction Filtering: Order-Selective Spectral Regularization for Multimodal Humor Detection https://arxiv.org/abs/2611.12913 https://arxiv.org/pdf/2611.12913 https://arxiv.org/html/2611.12913
November 11, 2026 at 6:42 AM
Jiahao Qin, Yiwen Wang: M\"obius Interaction Filtering: Order-Selective Spectral Regularization for Multimodal Humor Detection https://arxiv.org/abs/2611.12913 https://arxiv.org/pdf/2611.12913 https://arxiv.org/html/2611.12913
Francisco Prieto-Castrillo, Set P\'erez-Gonz\'alez, Jes\'us S. P\'erez del R\'io
From Complex Regularization to Polynomial Flow: Capture and Escape in Rocking Dynamics
https://arxiv.org/abs/2611.13861
From Complex Regularization to Polynomial Flow: Capture and Escape in Rocking Dynamics
https://arxiv.org/abs/2611.13861
November 11, 2026 at 7:34 PM
Francisco Prieto-Castrillo, Set P\'erez-Gonz\'alez, Jes\'us S. P\'erez del R\'io
From Complex Regularization to Polynomial Flow: Capture and Escape in Rocking Dynamics
https://arxiv.org/abs/2611.13861
From Complex Regularization to Polynomial Flow: Capture and Escape in Rocking Dynamics
https://arxiv.org/abs/2611.13861
💡 Overfitting: the model is fitting noise in the training data rather than patterns that generalize. Keep the weights from the best validation epoch (early stopping), and try regularization or more data.
How would you tell real overfitting from a noisy bump?
How would you tell real overfitting from a noisy bump?
November 12, 2026 at 2:00 PM
💡 Overfitting: the model is fitting noise in the training data rather than patterns that generalize. Keep the weights from the best validation epoch (early stopping), and try regularization or more data.
How would you tell real overfitting from a noisy bump?
How would you tell real overfitting from a noisy bump?
Research shows vision-language models struggle with accuracy from visual distractions, even after identifying the relevant image. Implementing answer consistency regularization improves performance by an average of 16 percentage points, addressing this issue. https://arxiv.org/abs/2611.15039
Back to the Question: Tracing Visual Distraction Through Text Tokens
ArXiv link for Back to the Question: Tracing Visual Distraction Through Text Tokens
arxiv.org
November 14, 2026 at 2:00 PM
Research shows vision-language models struggle with accuracy from visual distractions, even after identifying the relevant image. Implementing answer consistency regularization improves performance by an average of 16 percentage points, addressing this issue. https://arxiv.org/abs/2611.15039
Research shows vision-language models struggle with accuracy around distractor images, even with correct target identification. With "answer consistency regularization," accuracy improves by 16 percentage points, reshaping how models ignore irrelevant data. https://arxiv.org/abs/2611.15039
Back to the Question: Tracing Visual Distraction Through Text Tokens
ArXiv link for Back to the Question: Tracing Visual Distraction Through Text Tokens
arxiv.org
November 13, 2026 at 9:20 PM
Research shows vision-language models struggle with accuracy around distractor images, even with correct target identification. With "answer consistency regularization," accuracy improves by 16 percentage points, reshaping how models ignore irrelevant data. https://arxiv.org/abs/2611.15039
Visual-language models struggle with accuracy due to distractor images. Answer consistency regularization increased circular accuracy by 16 points while maintaining single-image performance, marking a breakthrough in reducing distractions. https://arxiv.org/abs/2611.15039
Back to the Question: Tracing Visual Distraction Through Text Tokens
ArXiv link for Back to the Question: Tracing Visual Distraction Through Text Tokens
arxiv.org
November 13, 2026 at 8:50 PM
Visual-language models struggle with accuracy due to distractor images. Answer consistency regularization increased circular accuracy by 16 points while maintaining single-image performance, marking a breakthrough in reducing distractions. https://arxiv.org/abs/2611.15039
Research indicates vision-language models underperform with visual distractions, even when identifying target images. Implementing answer consistency regularization led to a 16-point accuracy boost while keeping single-image performance intact. https://arxiv.org/abs/2611.15039
Back to the Question: Tracing Visual Distraction Through Text Tokens
ArXiv link for Back to the Question: Tracing Visual Distraction Through Text Tokens
arxiv.org
November 13, 2026 at 8:40 PM
Research indicates vision-language models underperform with visual distractions, even when identifying target images. Implementing answer consistency regularization led to a 16-point accuracy boost while keeping single-image performance intact. https://arxiv.org/abs/2611.15039
Stanford researchers found vision-language models struggle with distractions even when they identify targets. Their method, answer consistency regularization, increases accuracy by 16 percentage points while reducing interference and enhancing VLM usability. https://arxiv.org/abs/2611.15039
Back to the Question: Tracing Visual Distraction Through Text Tokens
ArXiv link for Back to the Question: Tracing Visual Distraction Through Text Tokens
arxiv.org
November 13, 2026 at 10:30 PM
Stanford researchers found vision-language models struggle with distractions even when they identify targets. Their method, answer consistency regularization, increases accuracy by 16 percentage points while reducing interference and enhancing VLM usability. https://arxiv.org/abs/2611.15039