#SongZe
LightNews LightNews
arxiv-cs-cv.bsky.social
Bojun Yang, Haochen Zhou, Zhifang Zhang, Haobo Wang, Songze Li, Lei Feng
Purifying Backdoored Large Vision-Language Models by Removing Hijacked Directions
https://arxiv.org/abs/2610.09941
October 10, 2026 at 7:48 AM
cscv-bot.bsky.social
Bojun Yang, Haochen Zhou, Zhifang Zhang, Haobo Wang, Songze Li, Lei Feng: Purifying Backdoored Large Vision-Language Models by Removing Hijacked Directions https://arxiv.org/abs/2610.09941 https://arxiv.org/pdf/2610.09941 https://arxiv.org/html/2610.09941
October 8, 2026 at 6:41 AM
ultraeric.bsky.social
A street view on Songze Road

Shot on July 3rd, 2026

#shanghai
October 2, 2026 at 4:03 AM
csma-bot.bsky.social
Yapeng Li, Songze Li, Shuang Yu, Jing Yu, Zhixin Liu, Liqiang Wen, Tonghua Su: MASTraceBench: Diagnosing Collaboration Gains through Proposal Trajectories in LLM-Based Multi-Agent Systems https://arxiv.org/abs/2609.34496 https://arxiv.org/pdf/2609.34496 https://arxiv.org/html/2609.34496
September 29, 2026 at 6:42 AM
dailyrunway.bsky.social
chen duling for cosmopolitan china 33rd anniversary issue, photographed by tu songze
August 7, 2026 at 11:19 PM
nanoatp.bsky.social
[28/30] 187 Upvotes, 2 Comments, 2 Posts, arXiv:2607.17423

🆕TimeLens2: Generalist Video Temporal Grounding with Multimodal LLMs

Yuhan Zhu, Changlian Ma, Xiangyu Zeng, Xinhao Li, Zhiqiu Zhang, Songze Li
Video multimodal large language models (MLLMs) can describe what happens in a video, but rarely identify when the supporting evidence occurs. We study generalist video temporal grounding, in which one model predicts a variable-cardinality set of evidence intervals across video lengths, domains, query forms, and viewpoints. Existing training strategies are misaligned with this set-valued task: long-video labels often rely on brittle one-pass annotation, while reinforcement-learning rewards either fail to distinguish non-overlapping predictions or require fragile segment matching. TimeLens2 treats temporal evidence as an interval set throughout supervision and optimization. TimeLens2-93K constructs reliable multi-span supervision through caption-derived proposals, independent localization, cross-agent consensus, semantic verification, and boundary refinement. Our temporal Wasserstein reward computes exact one-dimensional \(W_1\) between uniform distributions over merged interval supports, providing dense, matching-free feedback under unequal cardinalities and equivalent fragmentation; temporal IoU complements it with precise-overlap feedback. Across seven benchmarks, TimeLens2-2B outperforms all size-matched baselines on every benchmark, while the 4B and 8B variants achieve state-of-the-art performance, surpassing open-source models with up to 397B parameters. The 2B, 4B, and 8B variants improve over their Qwen3-VL backbones by 14.2, 13.0, and 18.1 mIoU points, respectively.
July 24, 2026 at 12:04 AM
nanoatp.bsky.social
[29/30] 179 Upvotes, 2 Comments, 2 Posts, arXiv:2607.17423

🆕TimeLens2: Generalist Video Temporal Grounding with Multimodal LLMs

Yuhan Zhu, Changlian Ma, Xiangyu Zeng, Xinhao Li, Zhiqiu Zhang, Songze Li
Video multimodal large language models (MLLMs) can describe what happens in a video, but rarely identify when the supporting evidence occurs. We study generalist video temporal grounding, in which one model predicts a variable-cardinality set of evidence intervals across video lengths, domains, query forms, and viewpoints. Existing training strategies are misaligned with this set-valued task: long-video labels often rely on brittle one-pass annotation, while reinforcement-learning rewards either fail to distinguish non-overlapping predictions or require fragile segment matching. TimeLens2 treats temporal evidence as an interval set throughout supervision and optimization. TimeLens2-93K constructs reliable multi-span supervision through caption-derived proposals, independent localization, cross-agent consensus, semantic verification, and boundary refinement. Our temporal Wasserstein reward computes exact one-dimensional \(W_1\) between uniform distributions over merged interval supports, providing dense, matching-free feedback under unequal cardinalities and equivalent fragmentation; temporal IoU complements it with precise-overlap feedback. Across seven benchmarks, TimeLens2-2B outperforms all size-matched baselines on every benchmark, while the 4B and 8B variants achieve state-of-the-art performance, surpassing open-source models with up to 397B parameters. The 2B, 4B, and 8B variants improve over their Qwen3-VL backbones by 14.2, 13.0, and 18.1 mIoU points, respectively.
July 23, 2026 at 12:05 AM
oyatsumaster.bsky.social
2026年7月4日
#考古学のおやつ 1/2
中国、江蘇省（Jiangsu） 蘇州市（Suzhou） 呉中区（Wuzhong）・塘北遺跡（塘北遗址：Tangbei）の、崧沢文化（Songze）～西周（Western Zhou）時代・紀元前4000年～前700年の土器片141点を分析。
鉄分に富み石灰分を含まない在地の粘土を常に使用。混和材は崧沢文化では植物。馬橋文化（Maqiao culture）以降は、特に調理用の土器では砂の混和材が支配的。印文のある硬陶はシリカに富んだ胎土を用い、高温で焼成。→
July 4, 2026 at 1:04 PM
arxiv-cs-it.bsky.social
Zhenhao Huang, Kai Liang, Yuanming Shi, Songze Li, Youlong Wu
Asymptotically Optimal Secure Aggregation for Wireless Federated Learning with Multiple Servers
https://arxiv.org/abs/2506.23680
July 1, 2025 at 5:36 AM
arxiv-cs-it.bsky.social
Jinbao Zhu, Hengxuan Tang, Songze Li, Yijia Chang
Generalized Lagrange Coded Computing: A Flexible Computation-Communication Tradeoff for Resilient, Secure, and Private Computation
https://arxiv.org/abs/2204.11168
November 7, 2024 at 5:31 AM
csar-bot.bsky.social
Songze Liu, Hongkun Du, Shaowen Wang: Adaptive Cache Pollution Control for Large Language Model Inference Workloads Using Temporal CNN-Based Prediction and Priority-Aware Replacement https://arxiv.org/abs/2512.14151 https://arxiv.org/pdf/2512.14151 https://arxiv.org/html/2512.14151
December 17, 2025 at 6:29 AM
thehistocrat.bsky.social
Oh hey as part of my upcoming video project I made a little map of Middle Neolithic Chinese Cultures (approx. 5000-3000 BCE)
A map of middle neolithic China's archaeology cultures, including the Northern Hongshan, Yangshao and Dawenkou cultures, the Southern Central Daxi, Hemudu, Songze, Beiyinyangying and Xuejiagang cultures, and the far Southern Gaomiao, Dingsishan IV, Xiantouling and Keqiutou cultures. Finally in Taiwan we also see the Dabenkeng culture.
June 11, 2024 at 1:53 PM
cscv-bot.bsky.social
Tianxiang Jiang, Linquan Wu, Sheng Xia, Songze Li, Ziang Yan, Haoyu Yang, Yu Qiao, Yi Wang: Imagine Before You Predict: Interleaved Latent Visual Reasoning for Video Event Prediction https://arxiv.org/abs/2606.05769 https://arxiv.org/pdf/2606.05769 https://arxiv.org/html/2606.05769
June 5, 2026 at 6:40 AM
cscv-bot.bsky.social
Jiacheng Chen, Songze Li, Han Fu, Baoquan Zhao, Wei Liu, Yanyan Liang, Li Qing, Xudong Mao: Delta-Adapter: Scalable Exemplar-Based Image Editing with Single-Pair Supervision https://arxiv.org/abs/2605.07940 https://arxiv.org/pdf/2605.07940 https://arxiv.org/html/2605.07940
May 11, 2026 at 6:42 AM
cscv-bot.bsky.social
Yilong Huang, Songze Li: Beauty and the Beast: Imperceptible Perturbations Against Diffusion-Based Face Swapping via Directional Attribute Editing https://arxiv.org/abs/2601.22744 https://arxiv.org/pdf/2601.22744 https://arxiv.org/html/2601.22744
February 2, 2026 at 6:31 AM
cscv-bot.bsky.social
Hao Chen, Yiwei Wang, Songze Li: Bi-Erasing: A Bidirectional Framework for Concept Removal in Diffusion Models https://arxiv.org/abs/2512.13039 https://arxiv.org/pdf/2512.13039 https://arxiv.org/html/2512.13039
December 16, 2025 at 6:32 AM
cscv-bot.bsky.social
Songze Li, Mingyu Gao, Tonghua Su, Xu-Yao Zhang, Zhongjie Wang: Multimodal Continual Instruction Tuning with Dynamic Gradient Guidance https://arxiv.org/abs/2511.15164 https://arxiv.org/pdf/2511.15164 https://arxiv.org/html/2511.15164
November 20, 2025 at 6:30 AM
cscv-bot.bsky.social
Songze Li, Qixing Xu, Tonghua Su, Xu-Yao Zhang, Zhongjie Wang: Balancing Stability and Plasticity in Pretrained Detector: A Dual-Path Framework for Incremental Object Detection https://arxiv.org/abs/2504.10214 https://arxiv.org/pdf/2504.10214 https://arxiv.org/html/2504.10214
April 15, 2025 at 6:29 AM
cscv-bot.bsky.social
Songze Li, Tonghua Su, Xu-Yao Zhang, Qixing Xu, Zhongjie Wang: DUKAE: DUal-level Knowledge Accumulation and Ensemble for Pre-Trained Model-Based Continual Learning https://arxiv.org/abs/2504.06521 https://arxiv.org/pdf/2504.06521 https://arxiv.org/html/2504.06521
April 10, 2025 at 5:57 AM
cscv-bot.bsky.social
Chunxue Xu, Yiwei Wang, Bryan Hooi, Yujun Cai, Songze Li: How does Watermarking Affect Visual Language Models in Document Understanding? https://arxiv.org/abs/2504.01048 https://arxiv.org/pdf/2504.01048 https://arxiv.org/html/2504.01048
April 3, 2025 at 5:56 AM
chrdnet.bsky.social
#NewCitizensMovement activist #SongZe is serving 2.5 yrs for “picking quarrels”—a catch-all charge the #China govt uses to silence dissidents, journalists & lawyers.
👉 tenchu.org/pocd/public/...
August 18, 2025 at 12:24 PM
chrdnet.bsky.social
The appeal of New Citizens' Movement activist #Songze was rejected, upholding his 2.5-year sentence. The court had previously ruled the charge of "providing apps & tools to invade & unlawfully control computer info systems" unestablished.
Read more🗒️ www.nchrd.org/2023/09/song...
March 19, 2025 at 8:48 AM
arxiv-cs-cr.bsky.social
Chenqing Zhu, Yanbo Dai, Yulong Tian, Qingming Li, Songze Li
When the Aggregator Cheats: Data-Free Backdoors in Federated LLM-based QA Systems
https://arxiv.org/abs/2606.27511
June 29, 2026 at 8:10 PM
arxiv-cs-cr.bsky.social
Zehan Sun, Dingfan Chen, Songze Li
When Efficiency Backfires: Cascading LLMs Trigger Cascade Failure under Adversarial Attack
https://arxiv.org/abs/2605.17288
May 20, 2026 at 12:36 AM
arxiv-cs-cr.bsky.social
Xiaolei Zhang, Xiaojun Jia, Liquan Chen, Songze Li
CODE: A Contradiction-Based Deliberation Extension Framework for Overthinking Attacks on Retrieval-Augmented Generation
https://arxiv.org/abs/2601.13112
January 21, 2026 at 12:38 PM