Tootfinder

Jazzaria – Lost Insight
#acoustic

Critics of DEI often argue, that these programs promote hires purely on ethnicity, gender or sexual orientation while the qualification gets neglected.
If this was truely the case, then I would agree with them. These distinctive features should not play a role. However, the reality was (and apparently still is), that these factors do play a role and DEI measurements are an instrument to battle these, not to intensify them. The following example shows this best:
In 1942 - 1945 Nav…

Azure cost reduction/optimization for hosted applications using Artificial Intelligence. And it's REPEATABLE. Interested? Read on...
✅ Daniel Meppiel's Project Post
https://www.linkedin.com/posts/danielmeppiel_enterpriseai-ai-…

💥 What if AI could shrink your cloud bill by 80% in the time it takes to drink your coffee? | Daniel Meppiel
💥 What if AI could shrink your cloud bill by 80% in the time it takes to drink your coffee? 💥 I recently got it to happen—firsthand. What started as a quick experiment turned into something game-changing. Using VSCode's reusable prompts and the Azure MCP server, GitHub Copilot identified key optimizations for my app's infra in a predictable manner. Then, the Coding Agent implemented them. 📉 The result? An 80% smaller cloud bill. Just like that. In this 2-minute demo (linked below …

Quantifying Azure RBAC Wildcard Overreach
Christophe Parisel
https://arxiv.org/abs/2506.10755 https://arxiv.org/pdf/2506.10755…

Quantifying Azure RBAC Wildcard Overreach
Azure RBAC leverages wildcard permissions to simplify policy authoring, but this abstraction often obscures the actual set of allowed operations and undermines least-privilege guarantees. We introduce Belshazaar, a two-stage framework that targets both the effective permission set problem and the evaluation of wildcards permissions spread. First, we formalize Azure action syntax via a context free grammar and implement a compiler that expands any wildcard into its explicit action set. Second, w…

Broadband Spectral Modeling of Blazars: Constraining the Lorentz Factor and Gamma-Ray Emission Site
Ajay Sharma, Aishwarya Sarath, Sakshi Chaudhary, Debanjan Bose
https://arxiv.org/abs/2506.12832

Broadband Spectral Modeling of Blazars: Constraining the Lorentz Factor and Gamma-Ray Emission Site
We present a comprehensive temporal and spectral analysis of a few blazars using multi-wavelength observations. Rapid flux variations are quantified via the doubling/halving timescale method, revealing the shortest variability timescales of a few hours in $γ$-ray emissions. The broadband fractional variability is systematically computed and examined as a function of frequency, displaying a characteristic double-hump structure akin to the typical spectral energy distribution (SED) of blazars. T…

Welcome to "rowhome"!
As usual, you can count on real estate marketers to signal status shifts.
"Townhome," invented because "rowhouse" felt low-class, now feels low-class itself.
I think they would have done better to just say "rowhouse" and invoke the prestige of historic preservation.

Check out some renderings for Woodley Park’s Wardman Park!! (the former Marriott Wardman) - PoPville
Renderings are courtesy Wardman Park From a press release: "Carmel Partners announced today the start of pre-leasing for Aerie, the first residential tower to open at Wardman Park, a reimagined, 9.5-acre landmark property now home to a resort-style residential oasis. Set atop one of the city’s highest elevations in Woodley Park, Wardman Park is defined

SC-SOT: Conditioning the Decoder on Diarized Speaker Information for End-to-End Overlapped Speech Recognition
Yuta Hirano, Sakriani Sakti
https://arxiv.org/abs/2506.12672

SC-SOT: Conditioning the Decoder on Diarized Speaker Information for End-to-End Overlapped Speech Recognition
We propose Speaker-Conditioned Serialized Output Training (SC-SOT), an enhanced SOT-based training for E2E multi-talker ASR. We first probe how SOT handles overlapped speech, and we found the decoder performs implicit speaker separation. We hypothesize this implicit separation is often insufficient due to ambiguous acoustic cues in overlapping regions. To address this, SC-SOT explicitly conditions the decoder on speaker information, providing detailed information about "who spoke when". Specifi…

Jazzaria – Fire and Forget
#byncnd

This https://arxiv.org/abs/2409.15104 has been replaced.
initial toot: https://mastoxiv.page/@arXiv_csDC_…

PecSched: Preemptive and Efficient Cluster Scheduling for LLM Inference
The scaling of transformer-based Large Language Models (LLMs) has significantly expanded their context lengths, enabling applications where inputs exceed 100K tokens. Our analysis of a recent Azure LLM inference trace reveals a highly skewed long-tail distribution of input lengths, with approximately 80% of inputs shorter than 2K tokens. Long inputs constitute only a small fraction. Existing cluster-level LLM scheduling strategies, including First-In-First-Out (FIFO), reservation-based, and pri…

Mitigating Non-Target Speaker Bias in Guided Speaker Embedding
Shota Horiguchi, Takanori Ashihara, Marc Delcroix, Atsushi Ando, Naohiro Tawara
https://arxiv.org/abs/2506.12500

Mitigating Non-Target Speaker Bias in Guided Speaker Embedding
Obtaining high-quality speaker embeddings in multi-speaker conditions is crucial for many applications. A recently proposed guided speaker embedding framework, which utilizes speech activities of target and non-target speakers as clues, drastically improved embeddings under severe overlap with small degradation in low-overlap cases. However, since extreme overlaps are rare in natural conversations, this degradation cannot be overlooked. This paper first reveals that the degradation is caused by…