Shengpeng and his bestie.

Sherwood talks about his return at SJHN YouTube!

Also, Reaves placed on IR, which opens the door for Sherwood's activation: www.youtube.com/@ShengPeng

"Nobody expected us to be in this position we're in, except for all of us."

Celebrini, Askarov & Chernyshov try to find the positive in a devastating #SJSharks loss.

Watch & subscribe at SJHN YouTube: https://www.youtube.com/@ShengPeng

WavChat: Surveyed existing spoken dialogue systems chronologically, categorized them, and reviewed core technologies, datasets, metrics, and benchmarks; identified limitations and future research directions.

Love how Collin Graf sums this up, winning hockey isn’t simply avoiding mistakes, it’s also how you react to mistakes.

Tough #SJSharks 7-4 loss vs. the Sens, full Graf, Brossoit & Warsofsky post-game interviews are up on SJHN YouTube: https://www.youtube.com/@ShengPeng

Read the study, 'When science meets geopolitics: global AI research network transformation (2000–2025)', by Marina Zhang et al in Science and Public Policy: bit.ly/4cGg984

"I loved watching it when I was a fan, but super-frustrating when you play against it."

Celebrini on McDavid' 5 points, also Dickinson & Nedeljkovic speak after a tough #SJSharks 5-2 loss to the Oilers.

All post-game interviews on SJHN YouTube: https://www.youtube.com/@ShengPeng

WavRAG, an audio-integrated RAG framework, directly processed raw audio for retrieval; it achieved comparable retrieval performance to ASR-Text RAG pipelines with a 10x speedup.

Discrete pitch tokens and target speaker prompts enhance expressive voice conversion; pretrains pitch VQVAE; uses masked pitch-conditioned flow matching for Mel-spectrogram synthesis; improves timbre similarity using global and time-varying timbre embeddings.

What's Jack Han's left field trade for #SJSharks?

He also likes how Sharks are slow-developing Dickinson & Misa.

Also, we talk Panarin trade & what Sherwood's extension might look like.

The San Jose Hockey Now Podcast is sponsored by Bring Hockey Back!

Watch/subscribe: www.youtube.com/@ShengPeng

S-DiT uses a sparse alignment algorithm to guide a latent diffusion transformer, improving robustness and naturalness; multi-condition classifier-free guidance for accent control; piecewise rectified flow for faster generation; achieved state-of-the-art zero-shot TTS quality.