Difference between revisions of "Resource:Seminar"

From MobiNetS
Jump to: navigation, search
 
(50 intermediate revisions by 3 users not shown)
Line 1: Line 1:
{{SemNote
{{SemNote
|time='''2025-03-07 10:30-12:00'''
|time='''2026-04-10 10:30'''
|addr=4th Research Building A518
|addr=4th Research Building A518
|note=Useful links: [[Resource:Reading_List|📚 Readling list]]; [[Resource:Seminar_schedules|📆 Schedules]]; [[Resource:Previous_Seminars|🧐 Previous seminars]].
|note=Useful links: [[Resource:Reading_List|📚 Readling list]]; [[Resource:Seminar_schedules|📆 Schedules]]; [[Resource:Previous_Seminars|🧐 Previous seminars]].
Line 8: Line 8:


{{Latest_seminar
{{Latest_seminar
|abstract = Large language models (LLMs) have demonstrated remarkable reasoning capabilities across diverse domains. Recent studies have shown that increasing test-time computation enhances LLMs' reasoning capabilities. This typically involves extensive sampling at inference time guided by an external LLM verifier, resulting in a two-player system. Despite external guidance, the effectiveness of this system demonstrates the potential of a single LLM to tackle complex tasks. Thus, we pose a new research problem: Can we internalize the searching capabilities to fundamentally enhance the reasoning abilities of a single LLM? This work explores an orthogonal direction focusing on post-training LLMs for autoregressive searching (i.e., an extended reasoning process with self-reflection and self-exploration of new strategies). To achieve this, we propose the Chain-of-Action-Thought (COAT) reasoning and a two-stage training paradigm: 1) a small-scale format tuning stage to internalize the COAT reasoning format and 2) a large-scale self-improvement stage leveraging reinforcement learning. Our approach results in Satori, a 7B LLM trained on open-source models and data. Extensive empirical evaluations demonstrate that Satori achieves state-of-the-art performance on mathematical reasoning benchmarks while exhibits strong generalization to out-of-domain tasks. Code, data, and models will be fully open-sourced.
|abstract = To effectively utilize heterogeneous specialized hardware units in modern GPUs, such as TensorCores and Tensor Memory Accelerators, this paper introduces PipeThreader, a new DNN compiler. PipeThreader proposes shifting scheduling functionality from hardware to software so as to enable more efficient and sophisticated computation pipelining with minimal manual effort. This is achieved through sTask-graph, a new DNN computation abstraction, a hierarchical hardware abstraction that captures the capabilities of specialized units, and new scheduling primitives. As a result, PipeThreader can discover efficient pipeline scheduling for well-studied DNN architectures like FlashAttention, achieving comparable or even superior performance. Additionally, it can uncover novel pipeline schemes for emerging models like Mamba2, delivering significantly better performance compared to state-of-the-art hand-crafted implementations. The code is open-sourced at https://github.com/tile-ai/tilelang.
|confname = Arxiv
|confname =OSDI'25
|link = https://arxiv.org/abs/2502.02508
|link = https://www.usenix.org/conference/osdi25/presentation/cheng
|title= Satori: Reinforcement Learning with Chain-of-Action-Thought Enhances LLM Reasoning via Autoregressive Search
|title= PipeThreader: Software-defined pipelining for efficient DNN execution
|speaker=Qinyong
|speaker=Junzhe
|date=2025-03-14
|date=2026-4-9
}}{{Latest_seminar
|abstract = Light bulbs have been recently explored to design Light Fidelity (LiFi) communication to battery-free tags, thus complementing Radiofrequency (RF) backscatter in the uplink. In this paper, we show that LiFi and RF backscatter are complementary and have unexplored interactions. We introduce PassiveLiFi, a battery-free system that uses LiFi to transmit RF backscatter at a meagre power budget. We address several challenges on the system design in the LiFi transmitter, the tag and the RF receiver. We design the first LiFi transmitter that implements a chirp spread spectrum (CSS) using the visible light spectrum. We use a small bank of solar cells for both communication and harvesting, and reconfigure them based on the amount of harvested energy and desired data rate. We further alleviate the low responsiveness of solar cells with a new low-power receiver design in the tag. We design and implement a novel technique for embedding multiple symbols in the RF backscatter based on delayed chirps. Experimental results with an RF carrier of 17dBm show that we can generate RF backscatter with a range of 92.1 meters/ μW consumed in the tag, which is almost double with respect to prior work.
|confname =ToN'23
|link = https://ieeexplore.ieee.org/document/10371205/
|title= LiFi for Low-Power and Long-Range RF Backscatter
|speaker=Mengyu
|date=2025-03-14
}}
}}


{{Resource:Previous_Seminars}}
{{Resource:Previous_Seminars}}

Latest revision as of 10:37, 10 April 2026

Time: 2026-04-10 10:30
Address: 4th Research Building A518
Useful links: 📚 Readling list; 📆 Schedules; 🧐 Previous seminars.

Latest

  1. [OSDI'25] PipeThreader: Software-defined pipelining for efficient DNN execution, Junzhe
    Abstract: To effectively utilize heterogeneous specialized hardware units in modern GPUs, such as TensorCores and Tensor Memory Accelerators, this paper introduces PipeThreader, a new DNN compiler. PipeThreader proposes shifting scheduling functionality from hardware to software so as to enable more efficient and sophisticated computation pipelining with minimal manual effort. This is achieved through sTask-graph, a new DNN computation abstraction, a hierarchical hardware abstraction that captures the capabilities of specialized units, and new scheduling primitives. As a result, PipeThreader can discover efficient pipeline scheduling for well-studied DNN architectures like FlashAttention, achieving comparable or even superior performance. Additionally, it can uncover novel pipeline schemes for emerging models like Mamba2, delivering significantly better performance compared to state-of-the-art hand-crafted implementations. The code is open-sourced at https://github.com/tile-ai/tilelang.

History

2024

2023

2022

2021

2020

  • [Topic] [ The path planning algorithm for multiple mobile edge servers in EdgeGO], Rong Cong, 2020-11-18

2019

2018

2017

Instructions

请使用Latest_seminar和Hist_seminar模板更新本页信息.

    • 修改时间和地点信息
    • 将当前latest seminar部分的code复制到这个页面中
    • 将{{Latest_seminar... 修改为 {{Hist_seminar...，并增加对应的日期信息|date=
    • 填入latest seminar各字段信息
    • link请务必不要留空，如果没有link则填本页地址 https://mobinets.org/index.php?title=Resource:Seminar
  • 格式说明
    • Latest_seminar:

{{Latest_seminar
|confname=
|link=
|title=
|speaker=
}}

    • Hist_seminar

{{Hist_seminar
|confname=
|link=
|title=
|speaker=
|date=
}}

Retrieved from "http://mobinets.cn/site/index.php?title=Resource:Seminar&oldid=3511"