Searcharxiv⌕ Search

arXiv subjects

Yunnan Wu

Publications and source records attributed to Yunnan Wu.

11 recordsLinked to original sources
arXiv2025

The Art of Tool Interface Design

We present an agentic framework, Thinker, which achieves state of art performance in challenging reasoning tasks for realistic customer service scenarios that involve complex business logic and human interactions via long horizons. On the $τ$-bench retail dataset, Thinker achieves 82.6\% success rate with GPT-4o (version 2024-06-01) (baseline: 68.3\%), and 81.9\% success rate with Llama-3.1 405B (baseline: 49.6\%), without any fine-tuning. Thinker effectively closes the gap in reasoning capabilities between the base models by introducing proper structure. The key features of the Thinker framework are: (1) State-Machine Augmented Generation (SMAG), which represents business logic as state machines and the LLM uses state machines as tools. (2) Delegation of tasks from the main reasoning loop to LLM-powered tools. (3) Adaptive context management. Our prompting-only solution achieves signficant gains, while still maintaining a standard agentic architecture with a ReAct style reasoning loop. The key is to innovate on the tool interface design, as exemplified by SMAG and the LLM-powered tools.

cs.AI↗
arXiv2024

Scaling User Modeling: Large-scale Online User Representations for Ads Personalization in Meta

Effective user representations are pivotal in personalized advertising. However, stringent constraints on training throughput, serving latency, and memory, often limit the complexity and input feature set of online ads ranking models. This challenge is magnified in extensive systems like Meta's, which encompass hundreds of models with diverse specifications, rendering the tailoring of user representation learning for each model impractical. To address these challenges, we present Scaling User Modeling (SUM), a framework widely deployed in Meta's ads ranking system, designed to facilitate efficient and scalable sharing of online user representation across hundreds of ads models. SUM leverages a few designated upstream user models to synthesize user embeddings from massive amounts of user features with advanced modeling techniques. These embeddings then serve as inputs to downstream online ads ranking models, promoting efficient representation sharing. To adapt to the dynamic nature of user features and ensure embedding freshness, we designed SUM Online Asynchronous Platform (SOAP), a latency free online serving system complemented with model freshness and embedding stabilization, which enables frequent user model updates and online inference of user embeddings upon each user request. We share our hands-on deployment experiences for the SUM framework and validate its superiority through comprehensive experiments. To date, SUM has been launched to hundreds of ads ranking models in Meta, processing hundreds of billions of user requests daily, yielding significant online metric gains and improved infrastructure efficiency.

cs.IR↗
arXiv2021

$AIR^2$ for Interaction Prediction

The 2021 Waymo Interaction Prediction Challenge introduced a problem of predicting the future trajectories and confidences of two interacting agents jointly. We developed a solution that takes an anchored marginal motion prediction model with rasterization and augments it to model agent interaction. We do this by predicting the joint confidences using a rasterized image that highlights the ego agent and the interacting agent. Our solution operates on the cartesian product space of the anchors; hence the $"^2"$ in $AIR^2$. Our model achieved the highest mAP (the primary metric) on the leaderboard.

cs.CV↗
arXiv2021

QK Iteration: A Self-Supervised Representation Learning Algorithm for Image Similarity

Self-supervised representation learning is a fundamental problem in computer vision with many useful applications (e.g., image search, instance level recognition, copy detection). In this paper we present a new contrastive self-supervised representation learning algorithm in the context of Copy Detection in the 2021 Image Similarity Challenge hosted by Facebook AI Research. Previous work in contrastive self-supervised learning has identified the importance of being able to optimize representations while ``pushing'' against a large number of negative examples. Representative previous solutions either use large batches enabled by modern distributed training systems or maintain queues or memory banks holding recently evaluated representations while relaxing some consistency properties. We approach this problem from a new angle: We directly learn a query model and a key model jointly and push representations against a very large number (e.g., 1 million) of negative representations in each SGD step. We achieve this by freezing the backbone on one side and by alternating between a Q-optimization step and a K-optimization step. During the competition timeframe, our algorithms achieved a micro-AP score of 0.3401 on the Phase 1 leaderboard, significantly improving over the baseline $μ$AP of 0.1556. On the final Phase 2 leaderboard, our model scored 0.1919, while the baseline scored 0.0526. Continued training yielded further improvement. We conducted an empirical study to compare the proposed approach with a SimCLR style strategy where the negative examples are taken from the batch only. We found that our method ($μ$AP of 0.3403) significantly outperforms this SimCLR-style baseline ($μ$AP of 0.2001).

cs.CV↗
arXiv2010

A Survey on Network Codes for Distributed Storage

Distributed storage systems often introduce redundancy to increase reliability. When coding is used, the repair problem arises: if a node storing encoded information fails, in order to maintain the same level of reliability we need to create encoded information at a new node. This amounts to a partial recovery of the code, whereas conventional erasure coding focuses on the complete recovery of the information from a subset of encoded packets. The consideration of the repair network traffic gives rise to new design challenges. Recently, network coding techniques have been instrumental in addressing these challenges, establishing that maintenance bandwidth can be reduced by orders of magnitude compared to standard erasure codes. This paper provides an overview of the research results on this topic.

cs.IT↗
arXiv2009

Position Modulation Code for Rewriting Write-Once Memories

A write-once memory (wom) is a storage medium formed by a number of ``write-once'' bit positions (wits), where each wit initially is in a `0' state and can be changed to a `1' state irreversibly. Examples of write-once memories include SLC flash memories and optical disks. This paper presents a low complexity coding scheme for rewriting such write-once memories, which is applicable to general problem configurations. The proposed scheme is called the \emph{position modulation code}, as it uses the positions of the zero symbols to encode some information. The proposed technique can achieve code rates higher than state-of-the-art practical solutions for some configurations. For instance, there is a position modulation code that can write 56 bits 10 times on 278 wits, achieving rate 2.01. In addition, the position modulation code is shown to achieve a rate at least half of the optimal rate.

cs.IT↗
arXiv2009

Congestion games with resource reuse and applications in spectrum sharing

In this paper we consider an extension to the classical definition of congestion games (CG) in which multiple users share the same set of resources and their payoff for using any resource is a function of the total number of users sharing it. The classical congestion games enjoy some very appealing properties, including the existence of a Nash equilibrium and that every improvement path is finite and leads to such a NE (also called the finite improvement property or FIP), which is also a local optimum to a potential function. On the other hand, this class of games does not model well the congestion or resource sharing in a wireless context, a prominent feature of which is spatial reuse. What this translates to in the context of a congestion game is that a users payoff for using a resource (interpreted as a channel) is a function of the its number of its interfering users sharing that channel, rather than the total number among all users. This makes the problem quite different. We will call this the congestion game with resource reuse (CG-RR). In this paper we study intrinsic properties of such a game; in particular, we seek to address under what conditions on the underlying network this game possesses the FIP or NE. We also discuss the implications of these results when applied to wireless spectrum sharing

cs.GT↗
arXiv2009

A Construction of Systematic MDS Codes with Minimum Repair Bandwidth

In a distributed storage system based on erasure coding, an important problem is the \emph{repair problem}: If a node storing a coded piece fails, in order to maintain the same level of reliability, we need to create a new encoded piece and store it at a new node. This paper presents a construction of systematic $(n,k)$-MDS codes for $2k\le n$ that achieves the minimum repair bandwidth when repairing from $k+1$ nodes.

cs.IT↗
arXiv2008

Network Coding for Distributed Storage Systems

Distributed storage systems provide reliable access to data through redundancy spread over individually unreliable nodes. Application scenarios include data centers, peer-to-peer storage systems, and storage in wireless networks. Storing data using an erasure code, in fragments spread across nodes, requires less redundancy than simple replication for the same level of reliability. However, since fragments must be periodically replaced as nodes fail, a key question is how to generate encoded fragments in a distributed way while transferring as little data as possible across the network. For an erasure coded system, a common practice to repair from a node failure is for a new node to download subsets of data stored at a number of surviving nodes, reconstruct a lost coded block using the downloaded data, and store it at the new node. We show that this procedure is sub-optimal. We introduce the notion of regenerating codes, which allow a new node to download \emph{functions} of the stored data from the surviving nodes. We show that regenerating codes can significantly reduce the repair bandwidth. Further, we show that there is a fundamental tradeoff between storage and repair bandwidth which we theoretically characterize using flow arguments on an appropriately constructed graph. By invoking constructive results in network coding, we introduce regenerating codes that can achieve any point in this optimal tradeoff.

cs.NI↗
arXiv2005

A low-cost time-hopping impulse radio system for high data rate transmission

We present an efficient, low-cost implementation of time-hopping impulse radio that fulfills the spectral mask mandated by the FCC and is suitable for high-data-rate, short-range communications. Key features are: (i) all-baseband implementation that obviates the need for passband components, (ii) symbol-rate (not chip rate) sampling, A/D conversion, and digital signal processing, (iii) fast acquisition due to novel search algorithms, (iv) spectral shaping that can be adapted to accommodate different spectrum regulations and interference environments. Computer simulations show that this system can provide 110Mbit/s at 7-10m distance, as well as higher data rates at shorter distances under FCC emissions limits. Due to the spreading concept of time-hopping impulse radio, the system can sustain multiple simultaneous users, and can suppress narrowband interference effectively.

cs.IT↗