arXiv subjects

Shwan Ashrafi

Publications and source records attributed to Shwan Ashrafi.

5 recordsLinked to original sources

arXiv2026

The Crowded Embedding Space: A Mean-Field Mechanism for Emergent Marginalization in Retrieval-Augmented Agents

Retrieval-augmented generative agents rely on retrieval for grounding, yet are typically evaluated on a query-by-query basis. This isolates interactions that are geometrically coupled in a shared embedding space. For example, we show that the high document density required to serve majority interests (e.g., generic "Crime" movies) can geometrically overcrowd the retrieval neighborhood of a semantically similar minority (e.g., "Film Noir"), effectively expelling minority content from top-$k$ results. We introduce a formal framework to analyze how such goal collisions in dense retrieval induce fundamental performance limits and emergent fairness issues inherent to spatial crowding. In our static analysis, we demonstrate that for a fixed embedding space, a phase transition occurs where minority user goals suffer a catastrophic collapse in performance as the density of majority goals increases. We then extend this to a dynamic model and derive a non-linear Fokker-Planck equation that governs the evolution of document embeddings as the agent updates them to maximize retrieval accuracy. Our analysis reveals that this local relevance objective triggers an emergent global mechanism that systematically marginalizes minority interests. We prove that such objectives drive the system to self-organize into a state that exclusively serves majority interests. These results provide a theoretical foundation for understanding a critical grounding failure mode in retrieval-augmented agents.

cs.IR↗

arXiv2026

Budget-Aware Anytime Reasoning with LLM-Synthesized Preference Data

We study the reasoning behavior of large language models (LLMs) under limited computation budgets. In such settings, producing useful partial solutions quickly is often more practical than exhaustive reasoning, which incurs high inference costs. Many real-world tasks, such as trip planning, require models to deliver the best possible output within a fixed reasoning budget. We introduce an anytime reasoning framework and the Anytime Index, a metric that quantifies how effectively solution quality improves as reasoning tokens increase. To further enhance efficiency, we propose an inference-time self-improvement method using LLM-synthesized preference data, where models learn from their own reasoning comparisons to produce better intermediate solutions. Experiments on NaturalPlan (Trip), AIME, and GPQA datasets show consistent gains across Grok-3, GPT-oss, GPT-4.1/4o, and LLaMA models, improving both reasoning quality and efficiency under budget constraints.

cs.CL↗

arXiv2018

Large-Scale Video Classification with Feature Space Augmentation coupled with Learned Label Relations and Ensembling

This paper presents the Axon AI's solution to the 2nd YouTube-8M Video Understanding Challenge, achieving the final global average precision (GAP) of 88.733% on the private test set (ranked 3rd among 394 teams, not considering the model size constraint), and 87.287% using a model that meets size requirement. Two sets of 7 individual models belonging to 3 different families were trained separately. Then, the inference results on a training data were aggregated from these multiple models and fed to train a compact model that meets the model size requirement. In order to further improve performance we explored and employed data over/sub-sampling in feature space, an additional regularization term during training exploiting label relationship, and learned weights for ensembling different individual models.

cs.CV↗

arXiv2016

Performance Analysis of CSMA with Multi-Packet Reception: The Inhomogeneous Case

The problem of Carrier Sense Multiple Access (CSMA) with multi-packet reception (MPR) is studied. Most prior work has focused on the homogeneous case, where all the mobile users are assumed to have identical packet arrival rates and transmission probabilities. The inhomogeneous case remains largely open in the literature. In this work, we make a first step towards this open problem by deriving throughput and delay expressions for inhomogeneous CSMA, with a particular focus on a family of MPR models called the "all-or-nothing" symmetric MPR. This family of MPR models allows us to overcome several technical challenges associated with conventional analysis and to derive accurate throughput and delay expressions in the large-systems regime. Interestingly, this family of MPR models is still general enough to include a number of useful MPR techniques - such as successive interference cancellation (SIC), compute-and-forward (C&F), and successive compute-and-forward (SCF) - as special cases. Based on these throughput and delay expressions, we provide theoretical guidelines for meeting quality-of-service requirements and for achieving global stability; we also evaluate the performances of various MPR techniques, highlighting the clear advantages offered by SCF.

cs.IT↗

arXiv2014

On Dissemination Time of Random Linear Network Coding in Ad-hoc Networks

Random linear network coding (RLNC) unicast protocol is analyzed over a rapidly-changing network topology. We model the probability mass function (pmf) of the dissemination time as a sequence of independent geometric random variables whose success probability changes with every successful reception of an innovative packet. We derive a tight approximation of the average networked innovation probability conditioned on network dimension increase. We show through simulations that our approximations for the average dissemination time and its pmf are tight. We then propose to use a RLNC-based broadcast dissemination protocol over a general dynamic topology where nodes are chosen for transmission based on average innovative information that they can provided to the rest of the network. Simulation results show that information disseminates considerably faster as opposed to standard RLNC algorithm where nodes are chosen uniformly at random.

cs.IT↗