AutoSOTA

Redefining SOTA. Bringing Research Back to What Matters.

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper ARC Is a Vision Problem! , AutoSOTA explored a follow-up enhancement and improved pass@1 to 0.600 (+0.200) using edge-aware loss weighting and longer test-time-training optimization. Huge thanks to the authors for the excellent work and for opening up such a valuable research direction.
AutoSOTA Project: https://t.co/w3XcHD6Jua
Enhancement Details: https://t.co/yPUAViYrUz
Original paper: https://t.co/u31TO7xVg8

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper IsoCLIP: Decomposing CLIP Projectors for Efficient Intra-modal Alignment, AutoSOTA explored a follow-up enhancement and improved mAP to 27.39 (+1.33%) using soft spectral thresholding and multi-band similarity ensemble. We would be grateful for any feedback from the authors and the community.
AutoSOTA Project: https://t.co/w3XcHD6Jua
Enhancement Details: https://t.co/aThRiF9t4s
Original paper: https://t.co/YggQ58YQnp

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper Mirror Illusion Art, AutoSOTA explored a follow-up enhancement and improved shape_score to 0.1818 (+5.4%) using checkpoint parameter evaluation and a longer shape optimization phase. Huge thanks to the authors for the excellent work and for opening up such a valuable research direction.
AutoSOTA Project: https://t.co/w3XcHD6Jua
Enhancement Details: https://t.co/PIuYgw4qp4
Original paper: https://t.co/XzZu4nHRSV

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Following the latest CVPR paper SAMTok: Representing Any Mask with Two Words by @xtl994 , AutoSOTA built on this inspiring work and further improved gIoU to 82.4 (+0.7) using higher MLLM image resolution from 448 to 896. We are grateful to the authors for the strong foundation that made this follow-up possible.
AutoSOTA Project: https://t.co/w3XcHD7hjI
Enhancement Details: https://t.co/t0REzJoRpB
Original paper: https://t.co/Q4QfCmdfwj

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by latest CVPR paper Statistical Characteristic-Guided Denoising for Rapid High-Resolution Transmission Electron Microscopy Imaging, AutoSOTA reproduced and extended the result, improving PSNR to 27.0843 (+2.01%) with four-way flip ensemble with original-weighted averaging. Grateful to the authors for the inspiring work that made this follow-up possible.
AutoSOTA Project: https://t.co/w3XcHD6Jua
Enhancement Details: https://t.co/oxr9MWuKkt
Original paper: https://t.co/NU3xmXmY3t

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper Mobile-VTON: High-Fidelity On-Device Virtual Try-On from @MingmingGong1 , AutoSOTA explored a follow-up enhancement and improved CLIP-I to 0.8783 (+5.16%) using weak guidance decay, horizontal flip ensemble, and timestep-aware garment weighting. We would be grateful for any feedback from the authors and the community.
AutoSOTA Project: https://t.co/w3XcHD6Jua
Enhancement Details: https://t.co/CscPZlJ2mv
Original paper: https://t.co/bbomUTSnaF

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper HiDRA: Hierarchical Degradation Representation and Adaptation with Generative Priors for Enhancing Infrared Vision, AutoSOTA explored a follow-up enhancement and improved LPIPS to 0.2911 (+9.1%) using unsharp mask post-processing with parameter grid search. Huge thanks to the authors for the excellent work and for opening up such a valuable research direction.
AutoSOTA Project: https://t.co/w3XcHD6Jua
Enhancement Details: https://t.co/ANjzDt74hG
Original paper: https://t.co/cm3Os1LeP1

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Following the latest CVPR paper Making Training-Free Diffusion Segmentors Scale with the Generative Power, AutoSOTA built on this inspiring work and further improved mIoU to 58.02% (+3.05 pp) using compound attention rescaling, max background handling, and flip TTA. We are grateful to the authors for the strong foundation that made this follow-up possible.
AutoSOTA Project: https://t.co/w3XcHD7hjI
Enhancement Details: https://t.co/iNDxzHwWqq
Original paper: https://t.co/Fb40LXKoEI

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by latest CVPR paper TR2M: Transferring Monocular Relative Depth to Metric Depth with Language Descriptions and Dual-Level Scale-Oriented Contrast, AutoSOTA reproduced and extended the result, improving delta1 to 0.944 (+0.001) with horizontal flip ensemble for depth prediction averaging. Grateful to the authors for the inspiring work that made this follow-up possible.
AutoSOTA Project: https://t.co/w3XcHD7hjI
Enhancement Details: https://t.co/YoQf8Au7M6
Original paper: https://t.co/hqP1QJqO54

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by latest CVPR paper BEV-SLD: Self-Supervised Scene Landmark Detection for Global Localization with LiDAR Bird's-Eye View Images, AutoSOTA reproduced and extended the result, improving SR to 100.00% (+1.69 pp) with a 5-run RANSAC ensemble that selects the solution with the highest inlier count. Grateful to the authors for the inspiring work that made this follow-up possible.
AutoSOTA Project: https://t.co/w3XcHD6Jua
Enhancement Details: https://t.co/xX4sQZirsB
Original paper: https://t.co/KlFIctYhge

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper KaLOS finds Consensus: A Meta-Algorithm for Evaluating Inter-Annotator Agreement in Complex Vision Tasks, AutoSOTA explored a follow-up enhancement and improved mean_alpha to 0.942825 (+16.71%) using IoU-centroid fusion and greedy ordering stabilization. We would be grateful for any feedback from the authors and the community.
AutoSOTA Project: https://t.co/w3XcHD7hjI
Enhancement Details: https://t.co/jfg7a7hnMG
Original paper: https://t.co/AsBzK3GH7M

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

We are grateful to the authors @dqj5182 for the strong foundation that made this follow-up possible. We hope this result can spark further discussion and ideas within the CVPR community.

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper Shoe Style-Invariant and Ground-Aware Learning for Dense Foot Contact Estimation, AutoSOTA explored a follow-up enhancement and improved contact F1 to 0.588 (+1.91%) using per-sample Otsu adaptive thresholding. We would be grateful for any feedback from the authors and the community.
AutoSOTA Project: https://t.co/w3XcHD7hjI
Original paper: https://t.co/9rcYOiZkFX

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Following the latest CVPR paper Deeper Thought, Weaker Aim: Understanding and Mitigating Perceptual Impairment during Reasoning in Multimodal Large Language Models, AutoSOTA built on this inspiring work and further improved ACC to 61.20% (+1.16 pp) using a binary visual attention mask with a larger focus neighborhood for VRGA. We are grateful to the authors for the strong foundation that made this follow-up possible.
AutoSOTA Project: https://t.co/w3XcHD6Jua
Enhancement Details: https://t.co/QOORGaxKCB
Original paper: https://t.co/lE3FP3gWdR

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

ChordEdit just received the CVPR Best Student Paper Honorable Mention yesterday, truly a very interesting piece of work!🎉

Inspired by it, we applied AutoS0TA for follow-up optimization from two directions. On the algorithm side, we introduced cleanup blending to better preserve background information from the source image, and used Prompt Similarity Auto-Tuning to dynamically adjust the editing strength based on the semantic similarity of the prompt. On the system side, we combined TF32 Tensor Core, SDPA / FlashAttention-2, and inference_mode to further improve inference efficiency.

As a result, PSNR improved from 23.02 to 25.11 (+9.1%), inference latency dropped to 0.25s/image (-32.4%), while CLIP-Whole remained at 97.7% of the baseline. The two groups of optimizations are largely orthogonal: the quality gains come with almost no speed cost, and the speed gains come with almost no quality degradation.

AutoSOTA Project: https://t.co/w3XcHD7hjI
Enhancement Details: https://t.co/GPjVzj909s
Original paper: https://t.co/XbSU16m5GM

We warmly welcome feedback and discussions from the ChordEdit authors @LiangsiLu and the broader community!

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper FocusUI: Efficient UI Grounding via Position-Preserving Visual Token Selection, AutoSOTA explored a follow-up enhancement and improved All-avg to 41.75% (+2.03%) using confidence-weighted multi-scale TTA and dual-threshold region merging. Huge thanks to the authors for the excellent work and for opening up such a valuable research direction.
AutoSOTA Project: https://t.co/w3XcHD7hjI
Enchantement Detail: https://t.co/NyrBaElEuc
Original paper: https://t.co/C2xTd8NIYz

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Following the latest CVPR paper Revisiting F-measure Optimization in Multi-Label Classification: A Sampling-based Approach, AutoSOTA built on this inspiring work and further improved Instance F1 to 0.5655 (+5.3%) using more Monte Carlo samples, a lower sampling temperature, and multi-run P-matrix averaging. We are grateful to the authors for the strong foundation that made this follow-up possible.
AutoSOTA Project: https://t.co/w3XcHD7hjI
Enchantement Detail: https://t.co/03AbysBNhG
Original paper: https://t.co/hl0B2e1M9g

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper GeoMotion: Rethinking Motion Segmentation via Latent 4D Geometry, AutoSOTA explored a follow-up enhancement and improved J&F to 0.86925 (+1.02%) using horizontal flip test-time augmentation with averaged original and flipped predictions. We would be grateful for any feedback from the authors and the community.
AutoSOTA Project: https://t.co/3qoyNu7L8x
Original paper: https://t.co/okmFf0qPHj
We hope this result can spark further discussion and ideas within the CVPR community. We would be grateful for any feedback or suggestions. Thank you! @deblinaforAI

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by the latest CVPR paper Rethinking SNN Online Training and Deployment: Gradient-Coherent Learning via Hybrid-Driven LIF Model, AutoSOTA explored a follow-up enhancement and improved CIFAR-100 Top-1 to 78.93% (+0.14 pp) using ECA channel attention for lightweight channel-wise recalibration on spiking ResNet features. Huge thanks to the authors for the excellent work and for opening up such a valuable research direction.
AutoSOTA Project: https://t.co/EEcsIjizRp
Original paper: https://t.co/InAGCkgAnB
We hope this result can spark further discussion and ideas within the CVPR community. We would be grateful for any feedback or suggestions. Thank you! @anfurnari

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Following the latest CVPR paper Generalizable Knowledge Distillation from Vision Foundation Models for Semantic Segmentation, AutoSOTA built on this inspiring work and further improved Cityscapes mIoU to 54.72% (+2.70 pp) using multi-scale logit averaging at 0.75×, 1.0×, and 1.25× scales. We are grateful to the authors for the strong foundation that made this follow-up possible.
AutoSOTA Project: https://t.co/SHs7d4mfx1
Original paper: https://t.co/LDHSREW6zu
We hope this result can spark further discussion and ideas within the CVPR community. We would be grateful for any feedback or suggestions. Thank you! @CSProfKGD

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation

[AutoSOTA @AutoSOTA11]

Inspired by @oyaceliktutan 's latest CVPR paper Unified Number-Free Text-to-Motion Generation Via Flow Matching, AutoSOTA reproduced and extended the result, improving R-precision@1 to 0.09777 (+6.5%) with time-dependent CFG scheduling with stronger text conditioning early and relaxed guidance later. Grateful to the authors for the inspiring work that made this follow-up possible.
AutoSOTA Project: https://t.co/hfS6fVzzO7
Original paper: https://t.co/CzQnceEwF1
We hope this result can spark further discussion and ideas within the CVPR community. We would be grateful for any feedback or suggestions. Thank you! @YVinker

#AutoSOTA_LiveCVPR #CVPR2026 #AutoSOTA #AIScientist #AutoResearch #MachineLearning #AcademicX #ResearchAutomation