arXiv subjects

Yu-Jung Chen

Publications and source records attributed to Yu-Jung Chen.

5 recordsLinked to original sources

arXiv2026

Molecule-specific diffusion and desorption of interstellar ices on carbonaceous dust

Interstellar ices form on dust grains in the coldest regions of molecular clouds and preserve key volatile reservoirs that could be incorporated into protoplanetary disks during star and planet formation. However, the effect of the dust surface composition on the ice structure and spectroscopic behavior remains poorly constrained. We present a comparative laboratory study of astrophysically relevant ices (CO, CO2, and H2O) deposited on inert calcium fluoride (CaF2) substrate and carbonaceous dust analogs under interstellar conditions. Infrared spectroscopy and temperature-programmed desorption reveal pronounced molecule-specific infrared spectral responses to the amorphous carbonaceous surface. CO and CO2 both exhibit broadened absorption bands, redshifted band positions, and delayed desorption, arising from thermally activated diffusion into the porous dust matrix and indicating strong molecule-surface interactions. By contrast, H2O varies only very little spectrally and thermally, indicating weak wetting and limited coupling to the substrate. These results provide direct laboratory evidence that dust-ice interfaces can affect the ice structure and desorption kinetics of interstellar ices even in thick ice layers. These findings offer new constraints for interpreting infrared absorption bands in astronomical observations and highlight the importance of surface effects in models of interstellar ice chemistry.

astro-ph.GA↗

arXiv2026

Guess-Verify-Refine: Data-Aware Top-K for Sparse-Attention Decoding on Blackwell via Temporal Correlation

Sparse-attention decoders rely on exact Top-K selection to choose the most important key-value entries for each query token. In long-context LLM serving, this Top-K stage runs once per decode query and becomes a meaningful latency bottleneck even when the indexer and attention kernels are already highly optimized. We present \textbf{Guess-Verify-Refine (GVR)}, a data-aware exact Top-K algorithm for sparse-attention decoding on NVIDIA Blackwell. GVR exploits temporal correlation across consecutive decode steps: it uses the previous step's Top-K as a prediction signal, computes pre-indexed statistics, narrows to a valid threshold by secant-style counting in 1-2 global passes, verifies candidates with a ballot-free collector, and finishes exact selection in shared memory. We connect this behavior to the Toeplitz / RoPE structure of DeepSeek Sparse Attention (DSA) indexer scores and validate the design on real DeepSeek-V3.2 workloads integrated into TensorRT-LLM. GVR achieves an average \textbf{1.88x} single-operator speedup over the production radix-select kernel, with up to \textbf{2.42x} per layer per step, while preserving bit-exact Top-K outputs. In controlled TEP8 min-latency deployment, it improves end-to-end TPOT by up to \textbf{7.52%} at 100K context, with larger gains at longer contexts and smaller but still positive gains under speculative decoding. While implemented and validated in the current TensorRT-LLM DSA stack on Blackwell, the same principle may extend to sparse-attention decoders whose decode-phase Top-K exhibits temporal stability.

cs.DC↗

arXiv2026

AVO: Agentic Variation Operators for Autonomous Evolutionary Search

Agentic Variation Operators (AVO) are a new family of evolutionary variation operators that replace the fixed mutation, crossover, and hand-designed heuristics of classical evolutionary search with autonomous coding agents. Rather than confining a language model to candidate generation within a prescribed pipeline, AVO instantiates variation as a self-directed agent loop that can consult the current lineage, a domain-specific knowledge base, and execution feedback to propose, repair, critique, and verify implementation edits. We evaluate AVO on attention, among the most aggressively optimized kernel targets in AI, on NVIDIA Blackwell (B200) GPUs. Over 7 days of continuous autonomous evolution on multi-head attention, AVO discovers kernels that outperform cuDNN by up to 3.5% and FlashAttention-4 by up to 10.5% across the evaluated configurations. The discovered optimizations transfer readily to grouped-query attention, requiring only 30 minutes of additional autonomous adaptation and yielding gains of up to 7.0% over cuDNN and 9.3% over FlashAttention-4. Together, these results show that agentic variation operators move beyond prior LLM-in-the-loop evolutionary pipelines by elevating the agent from candidate generator to variation operator, and can discover performance-critical micro-architectural optimizations that produce kernels surpassing state-of-the-art expert-engineered attention implementations on today's most advanced GPU hardware.

cs.LG↗

arXiv2017

New measurements on water ice photodesorption and product formation under ultraviolet irradiation

The photodesorption of icy grain mantles has been claimed to be responsible for the abundance of gas-phase molecules toward cold regions. Being water a ubiquitous molecule, it is crucial to understand its role in photochemistry and its behavior under an ultraviolet field. We report new measurements on the UV-photodesorption of water ice and its H$_2$, OH, and O$_2$ photoproducts using a calibrated quadrupole mass spectrometer. Solid water was deposited under ultra-high-vacuum conditions and then UV-irradiated at various temperatures starting from 8 K with a microwave discharged hydrogen lamp. Deuterated water was used for confirmation of the results. We found a photodesorption yield of 1.3 $\times$ 10$^{-3}$ molecules per incident photon for water, and 0.7 $\times$ 10$^{-3}$ molecules per incident photon for deuterated water at the lowest irradiation temperature, 8 K. The photodesorption yield per absorbed photon is given and comparison with astrophysical scenarios, where water ice photodesorption could account for the presence of gas-phase water toward cold regions in the absence of a thermal desorption process is addressed.

astro-ph.IM↗

arXiv2016

868 MHz Wireless Sensor Network - A Study

Today 2.4 GHz based wireless sensor networks are increasing at a tremendous pace, and are seen in widespread applications. Product innovation and support by many vendors in 2.4 GHz makes it a preferred choice, but the networks are prone to issues like interference, and range issues. On the other hand, the less popular 868 MHz in the ISM band has not seen significant usage. In this paper we explore the use of 868 MHz channel to implement a wireless sensor network, and study the efficacy of this channel

cs.NI↗