Data Synthesis and Parameter-Efficient Fine-Tuning for Low-Resource NMT: A Case Study on Q'eqchi' Mayan
この研究では、低リソース言語NMTのために、データ合成方法を開発しました。これにより、データ合成されたコーパスを使用し、NMTモデルをパラメータ効率的にフィーヌチュン化できます。
- 用途
- NMT低リソースデータ合成
- 難易度
- Hard
- コスト
- Medium
「LoRA」の検索結果
77 件この研究では、低リソース言語NMTのために、データ合成方法を開発しました。これにより、データ合成されたコーパスを使用し、NMTモデルをパラメータ効率的にフィーヌチュン化できます。
.deep learningモデルの重み操作を可能にするツールであるBrainSurgeryを提案し、層の再構成、精度のキャスティング、低ランク分解、構造的デバッグの実行を示します。
Contextual queueing bandits provide a framework for learning to schedule heterogeneous jobs under unknown cont
When running marketing campaigns, retailers must decide which products to promote and which users to target. T
この研究では、機械学習モデルをプライバシー保護のための適応化する際、プライバシー保護の実験的な効果を分析することに関与します。
この研究では、強力な防御ガードモデルと低パラメータのLLMを組み合わせたデュアルモデルシステムを導入し、安全なLLMのデプロイに使用できます。
Understanding tactical organisation of association football, hereafter referred to as football, requires ident
この論文では、data mining におけるビジュアルプログラミングフレームワーク、Orange Lab を提唱しました。これにより、Webベースのデータ分析環境を提供し、ユーザーフェイシングの分析ツールとしてデータ分
Recently, mobile edge computing (MEC)-enabled collaborative deep neural network (DNN) inference has emerged as
The rapid evolution of artificial intelligence has led to substantial advances in deep neural networks. Noneth
Process Reward Models (PRMs)は、セグメントごとにフィードバックを提供し、credit assignmentを改善します。しかし、我々はPRMsにおける偏微分に関する潜在的な偏りを識別し、この偏
Convolutions have successfully transitioned from image processing to the complex realm of non-Euclidean higher
A classical universal stabilization formula offers the practitioner no design freedom: it is a single, paramet
スライディングウィンドウのマルチアームバンドットでのパフォーマンスを最適化するために、アルゴリズムを提案します。
知識グラフの補完の評価を実現するために、新しいアプローチを提案します。
この論文では、エージェントの委譲能力を改善するために、新しいフレームワークを提案する。これにより、エージェントがより効率的にタスクを分割できる。
Spatial reasoning is a foundational capability for multimodal large language models (MLLMs) to perceive and op
Recursive self-design refers to AI-assisted modification of the mechanisms by which an AI system is built, eva
LLM間でモデル呼び出しと外部ツールの呼び出しが交互になり、サーバのサーヒングがステートレスの要求処理からステートフルなプログラム実行に移行します。これらのワークロードの評価は、各設計点ごとに専門的なアクセラレータ時間を
オーディオTokenと文書をモデル化するためにコーデックベースのARトークのジェネレーターが強力な文を音声の質を高めました。しかし、このアプローチでは、音声Tokenのシーケンスはテキストシーケンスより長くなるため、AR
Clinical diagnosis requires flexible use of multiple reasoning paradigms under incomplete patient information.
Semiconductor lithography inspection requires reliable detection of small pattern defects such as bridge, burr
この研究では、音楽生成における多様性を促進するためのオープンソース・フレームワークを開発します。このフレームワークは、音楽生成における多様性の促進を支援するために、進化的プロセスと多様性促進アルゴリズムを組み合わせたもの
この研究では、DragonflyアルゴリズムとTabu Searchを組み合わせたハイブリッドメタ ヘアスティックを提案し、Traveling Salesman問題を解決します。このアルゴリズムは、より早く、より正確に、
Writing Individualized Education Programs (IEPs) is a high-labor, knowledge-intensive document burden; English
Large language model agents increasingly rely on skills: reusable procedural documents encoding workflows, too
We present TruthSplit, an interactive system for multi-perspective argument analysis. Existing argumentation t
Understanding and reasoning over abstract visual content remains a challenge for current multi-modal large lan
The emergence of reasoning multimodal large language models (MLLMs), which generate explicit chain-of-thought
Dexterous manipulation presents substantial challenges for imitation learning due to its high-dimensional acti
Multi-modal Large Language Models (MLLMs) have achieved remarkable progress in video temporal grounding with r
4D generation (\textit{i.e.}, dynamic 3D generation) has recently emerged as a rapidly growing research fronti
Learning robot manipulation policies with deep neural networks from a single demonstration remains highly chal
Vision-Language-Action (VLA) models have become a powerful framework for robotic manipulation, and recent stud
Standard flow and diffusion pre-training matches the distribution of available data (e.g., molecules), which o
Reinforcement Learning (RL) has emerged as a pivotal post-training paradigm, yet it frequently suffers from un
We study data-driven real-time economic optimization of a multi-product chemical reactor when no reliable firs
Low-Rank Adaptation (LoRA) enables efficient federated fine-tuning of segmentation foundation models for medic
Exploratory manipulation often turns an apparent failed attempt into the key evidence for what to do next. For
While large language models (LLMs) offer promising reasoning capabilities, their integration into safety-criti
Reinforcement Learning from Human Feedback via Proximal Policy Optimization often suffers from policy mode col
Computational mental health research has predominantly centered on English-speaking populations, leaving Arabi
Large language models (LLMs) have shown promise in code summarization, yet their effectiveness for Hardware De
Object navigation requires a robot to search for an unobserved target in an unknown environment by deciding wh
Navigation using a monocular camera is pivotal for autonomous operation on tiny aerial robots due to their per
この論文では、VLAモデルをedgeハードウェアにデプロイするための手法を提案しています。この手法は、VLAモデルをedgeハードウェアにデプロイするためのフレームワークです。この手法は、edgeハードウェアを利用してV
Vision-Language-Action (VLA) models are emerging as a promising paradigm for robotic manipulation, enabling ge
Autonomous cyber-physical systems (CPS) sit at the intersection of Model-Based Systems Engineering (MBSE) and
Robots that operate over extended periods should not merely visit space; they should progressively understand
Two-sided matching markets often involve information that unfolds over time through interviews, repeated inter
オンライン学習でオフラインデータを使用できるように、線形バンディットのアルゴリズムを提案する。
LLMがプログラムを変換する際の動態学を調査し、変換は収束し、制限された吸引領域に収束することが多いことを示す。
この論文では、競合価格を考慮した需要モデル設計と、競合価格に対する意識のない学習とアルゴリズム的コラージュの間の関係を論じています。
Cooperative Multi-Agent Reinforcement Learning (MARL) frequently suffers from severe reward sparsity and explo
This paper proposes a two-stage pseudo anomaly-guided anomaly detection method (\textbf{T}wo-stage \textbf{P}s
Resource-constrained pricing controllers can make fixed-price inference impossible: the controller's resource
実用的なデータセット内の自然的実験の特定と分析と、それを用いた実験結果について論じます。
Agentic reinforcement learning (RL) enables LLM agents to improve continuously from environment rewards, yet t
Incentive design studies how a central authority can influence strategic agents through payments, subsidies, o
Deep learning has greatly advanced automatic speech recognition (ASR), enabling widespread deployment on edge
Multi-arm bandit algorithms are increasingly used in online platforms, clinical trials, and social science exp
Surrogate-assisted evolutionary algorithms (SAEAs) have been widely used for expensive black-box optimization
Particle Swarm Optimization (PSO) frequently suffers from premature convergence. This paper introduces a famil
ロボットの制御とパスの計画を高精度化するために、ロボットの制御アルゴリズムを開発する。
We study adversarial multi-armed bandits with and without delayed feedback under a safety-aware goal: achievin
この論文では、アノマリーディテクションにおける負の選択アルゴリズムを最適化するために、量子遺伝アルゴリズムを導入します。这により、検出器の生成効率を向上させることができ、精度も向上します。
Multimodal optimization requires finding many optima rather than merely keeping a diverse population. Yet most
We study whether simple algorithmic pricing systems can systematically produce collusive-like prices in multi-
We study a repeated information design setting in which the receiver, who is also the decision-maker, updates
大規模な不完全情報競争ゲーム(例:スターカフト、ドータ、カウンターストライク)における精査可能な均衡の取得は、sparse rewards と長時間探索のため計算上厄介になっています。この研究では、対手2人ゼロ和(2x0
Variational Quantum Algorithms (VQAs) are a leading approach to exploiting near-term quantum hardware, leverag
We decompose an evolutionary mixture-of-LoRA system on a from-scratch ~150M-parameter widened-D substrate (D=1
What happens when the strongest alliance member pressures a weaker member over territory and strategic control
Gradient-based preference optimization methods for large language model (LLM) alignment suffer from preference
Discovering governing equations from observational data remains a fundamental challenge in scientific modeling
This paper proposes Drain-Vortex Optimization (DVO), a population-based metaheuristic for continuous optimizat
Automatic heuristic design (AHD) has emerged as a promising paradigm for solving NP-hard combinatorial optimiz