The core of video understanding tasks, such as recognition, captioning, and tracking, is to automatically detect objects or actions in a video and analyze their temporal evolution. Despite sharing a common goal, different tasks often rely on distinct model architectures and annotation formats. In contrast, natural language processing benefits from a unified output space, i.e., text sequences, which simplifies the training of powerful foundational language models, such as GPT-3, with extensive training corpora. Inspired by this, we seek to unify the output space of video understanding tasks by using languages as labels and additionally introducing time and box tokens. In this way, a variety of video tasks could be formulated as video-grounded token generation. This enables us to address various types of video tasks, including classification (such as action recognition), captioning (covering clip captioning, video question answering, and dense video captioning), and localization tasks (such as visual object tracking) within a fully shared encoder-decoder architecture, following a generative framework. Through comprehensive experiments, we demonstrate such a simple and straightforward idea is quite effective and can achieve state-of-the-art or competitive results on seven video benchmarks, providing a novel perspective for more universal video understanding. Code is available at https://github.com/wangjk666/OmniVid.
Generative models are now widely used by graphic designers and artists. Prior works have shown that these models remember and often replicate content from their training data during generation. Hence as their proliferation increases, it has become important to perform a database search to determine whether the properties of the image are attributable to specific training data, every time before a generated image is used for professional purposes. Existing tools for this purpose focus on retrieving images of similar semantic content. Meanwhile, many artists are concerned with style replication in text-to-image models. We present a framework for understanding and extracting style descriptors from images. Our framework comprises a new dataset curated using the insight that style is a subjective property of an image that captures complex yet meaningful interactions of factors including but not limited to colors, textures, shapes, etc. We also propose a method to extract style descriptors that can be used to attribute style of a generated image to the images used in the training dataset of a text-to-image model. We showcase promising results in various style retrieval tasks. We also quantitatively and qualitatively analyze style attribution and matching in the Stable Diffusion model. Code and artifacts are available at https://github.com/learn2phoenix/CSD.
Can the rapid advances in code generation, function calling, and data analysis using large language models (LLMs) help automate the search and verification of hypotheses purely from a set of provided datasets? To evaluate this question, we present DiscoveryBench, the first comprehensive benchmark that formalizes the multi-step process of data-driven discovery. The benchmark is designed to systematically assess current model capabilities in discovery tasks and provide a useful resource for improving them. Our benchmark contains 264 tasks collected across 6 diverse domains, such as sociology and engineering, by manually deriving discovery workflows from published papers to approximate the real-world challenges faced by researchers, where each task is defined by a dataset, its metadata, and a discovery goal in natural language. We additionally provide 903 synthetic tasks to conduct controlled evaluations across task complexity. Furthermore, our structured formalism of data-driven discovery enables a facet-based evaluation that provides useful insights into different failure modes. We evaluate several popular LLM-based reasoning frameworks using both open and closed LLMs as baselines on DiscoveryBench and find that even the best system scores only 25%. Our benchmark, thus, illustrates the challenges in autonomous data-driven discovery and serves as a valuable resource for the community to make progress.
Large language models (LLMs) have demonstrated remarkable proficiency across various natural language processing (NLP) tasks. However, adapting LLMs to downstream applications requires computationally intensive and memory-demanding fine-tuning procedures. To alleviate these burdens, parameter-efficient fine-tuning (PEFT) techniques have emerged as a promising approach to tailor LLMs with minimal computational overhead. While PEFT methods offer substantial advantages, they do not fully address the pervasive issue of bias propagation from pre-training data. This work introduces Bias-Alleviating Low-Rank Adaptation (BA-LoRA), a novel PEFT method designed to counteract bias inheritance. BA-LoRA incorporates three distinct regularization terms: (1) a consistency regularizer, (2) a diversity regularizer, and (3) a singular value decomposition regularizer. These regularizers aim to enhance the models' consistency, diversity, and generalization capabilities during fine-tuning. We conduct extensive experiments on natural language understanding (NLU) and natural language generation (NLG) tasks using prominent LLMs such as LLaMA, Mistral, and Gemma. The results demonstrate that BA-LoRA outperforms LoRA and its state-of-the-art variants. Moreover, our method effectively mitigates the adverse effects of pre-training bias, leading to more reliable and robust model outputs. The code is available at https://github.com/cyp-jlu-ai/BA-LoRA.
Recent research on explainable recommendation generally frames the task as a standard text generation problem, and evaluates models simply based on the textual similarity between the predicted and ground-truth explanations. However, this approach fails to consider one crucial aspect of the systems: whether their outputs accurately reflect the users' (post-purchase) sentiments, i.e., whether and why they would like and/or dislike the recommended items. To shed light on this issue, we introduce new datasets and evaluation methods that focus on the users' sentiments. Specifically, we construct the datasets by explicitly extracting users' positive and negative opinions from their post-purchase reviews using an LLM, and propose to evaluate systems based on whether the generated explanations 1) align well with the users' sentiments, and 2) accurately identify both positive and negative opinions of users on the target items. We benchmark several recent models on our datasets and demonstrate that achieving strong performance on existing metrics does not ensure that the generated explanations align well with the users' sentiments. Lastly, we find that existing models can provide more sentiment-aware explanations when the users' (predicted) ratings for the target items are directly fed into the models as input. We will release our code and datasets upon acceptance.
Due to the singularity of real-world paired datasets and the complexity of low-light environments, this leads to supervised methods lacking a degree of scene generalisation. Meanwhile, limited by poor lighting and content guidance, existing zero-shot methods cannot handle unknown severe degradation well. To address this problem, we will propose a new zero-shot low-light enhancement method to compensate for the lack of light and structural information in the diffusion sampling process by effectively combining the wavelet and Fourier frequency domains to construct rich a priori information. The key to the inspiration comes from the similarity between the wavelet and Fourier frequency domains: both light and structure information are closely related to specific frequency domain regions, respectively. Therefore, by transferring the diffusion process to the wavelet low-frequency domain and combining the wavelet and Fourier frequency domains by continuously decomposing them in the inverse process, the constructed rich illumination prior is utilised to guide the image generation enhancement process. Sufficient experiments show that the framework is robust and effective in various scenarios. The code will be available at: \href{https://github.com/hejh8/Joint-Wavelet-and-Fourier-priors-guided-diffusion}{https://github.com/hejh8/Joint-Wavelet-and-Fourier-priors-guided-diffusion}.
Recent advancements in large language models (LLMs) have shown remarkable potential in automating machine learning tasks. However, existing LLM-based agents often struggle with low-diversity and suboptimal code generation. While recent work has introduced Monte Carlo Tree Search (MCTS) to address these issues, limitations persist in the quality and diversity of thoughts generated, as well as in the scalar value feedback mechanisms used for node selection. In this study, we introduce Introspective Monte Carlo Tree Search (I-MCTS), a novel approach that iteratively expands tree nodes through an introspective process that meticulously analyzes solutions and results from parent and sibling nodes. This facilitates a continuous refinement of the node in the search tree, thereby enhancing the overall decision-making process.Furthermore, we integrate a Large Language Model (LLM)-based value model to facilitate direct evaluation of each node's solution prior to conducting comprehensive computational rollouts. A hybrid rewarding mechanism is implemented to seamlessly transition the Q-value from LLM-estimated scores to actual performance scores. This allows higher-quality nodes to be traversed earlier.Applied to the various ML tasks, our approach demonstrates a6\% absolute improvement in performance compared to the strong open-source AutoML agents, showcasing its effectiveness in enhancing agentic AutoML systems.
Deep learning models lack shift invariance, making them sensitive to input shifts that cause changes in output. While recent techniques seek to address this for images, our findings show that these approaches fail to provide shift-invariance in time series, where the data generation mechanism is more challenging due to the interaction of low and high frequencies. Worse, they also decrease performance across several tasks. In this paper, we propose a novel differentiable bijective function that maps samples from their high-dimensional data manifold to another manifold of the same dimension, without any dimensional reduction. Our approach guarantees that samples -- when subjected to random shifts -- are mapped to a unique point in the manifold while preserving all task-relevant information without loss. We theoretically and empirically demonstrate that the proposed transformation guarantees shift-invariance in deep learning models without imposing any limits to the shift. Our experiments on six time series tasks with state-of-the-art methods show that our approach consistently improves the performance while enabling models to achieve complete shift-invariance without modifying or imposing restrictions on the model's topology. The source code is available on \href{https://github.com/eth-siplab/Shifting-the-Paradigm}{GitHub}.
While the current state-of-the-art dense retrieval models exhibit strong out-of-domain generalization, they might fail to capture nuanced domain-specific knowledge. In principle, fine-tuning these models for specialized retrieval tasks should yield higher effectiveness than relying on a one-size-fits-all model, but in practice, results can disappoint. We show that standard fine-tuning methods using an InfoNCE loss can unexpectedly degrade effectiveness rather than improve it, even for domain-specific scenarios. This holds true even when applying widely adopted techniques such as hard-negative mining and negative de-noising. To address this, we explore a training strategy that uses listwise distillation from a teacher cross-encoder, leveraging rich relevance signals to fine-tune the retriever. We further explore synthetic query generation using large language models. Through listwise distillation and training with a diverse set of queries ranging from natural user searches and factual claims to keyword-based queries, we achieve consistent effectiveness gains across multiple datasets. Our results also reveal that synthetic queries can rival human-written queries in training utility. However, we also identify limitations, particularly in the effectiveness of cross-encoder teachers as a bottleneck. We release our code and scripts to encourage further research.
Object pose estimation is a core means for robots to understand and interact with their environment. For this task, monocular category-level methods are attractive as they require only a single RGB camera. However, current methods rely on shape priors or CAD models of the intra-class known objects. We propose a diffusion-based monocular category-level 9D object pose generation method, MonoDiff9D. Our motivation is to leverage the probabilistic nature of diffusion models to alleviate the need for shape priors, CAD models, or depth sensors for intra-class unknown object pose estimation. We first estimate coarse depth via DINOv2 from the monocular image in a zero-shot manner and convert it into a point cloud. We then fuse the global features of the point cloud with the input image and use the fused features along with the encoded time step to condition MonoDiff9D. Finally, we design a transformer-based denoiser to recover the object pose from Gaussian noise. Extensive experiments on two popular benchmark datasets show that MonoDiff9D achieves state-of-the-art monocular category-level 9D object pose estimation accuracy without the need for shape priors or CAD models at any stage. Our code will be made public at https://github.com/CNJianLiu/MonoDiff9D.
Action segmentation is a core challenge in high-level video understanding, aiming to partition untrimmed videos into segments and assign each a label from a predefined action set. Existing methods primarily address single-person activities with fixed action sequences, overlooking multi-person scenarios. In this work, we pioneer textual reference-guided human action segmentation in multi-person settings, where a textual description specifies the target person for segmentation. We introduce the first dataset for Referring Human Action Segmentation, i.e., RHAS133, built from 133 movies and annotated with 137 fine-grained actions with 33h video data, together with textual descriptions for this new task. Benchmarking existing action recognition methods on RHAS133 using VLM-based feature extractors reveals limited performance and poor aggregation of visual cues for the target person. To address this, we propose a holistic-partial aware Fourier-conditioned diffusion framework, i.e., HopaDIFF, leveraging a novel cross-input gate attentional xLSTM to enhance holistic-partial long-range reasoning and a novel Fourier condition to introduce more fine-grained control to improve the action segmentation generation. HopaDIFF achieves state-of-the-art results on RHAS133 in diverse evaluation settings. The code is available at https://github.com/KPeng9510/HopaDIFF.git.
Recent progress in Large Vision-Language Models (LVLMs) has enabled promising applications in medical tasks, such as report generation and visual question answering. However, existing benchmarks focus mainly on the final diagnostic answer, offering limited insight into whether models engage in clinically meaningful reasoning. To address this, we present CheXStruct and CXReasonBench, a structured pipeline and benchmark built on the publicly available MIMIC-CXR-JPG dataset. CheXStruct automatically derives a sequence of intermediate reasoning steps directly from chest X-rays, such as segmenting anatomical regions, deriving anatomical landmarks and diagnostic measurements, computing diagnostic indices, and applying clinical thresholds. CXReasonBench leverages this pipeline to evaluate whether models can perform clinically valid reasoning steps and to what extent they can learn from structured guidance, enabling fine-grained and transparent assessment of diagnostic reasoning. The benchmark comprises 18,988 QA pairs across 12 diagnostic tasks and 1,200 cases, each paired with up to 4 visual inputs, and supports multi-path, multi-stage evaluation including visual grounding via anatomical region selection and diagnostic measurements. Even the strongest of 10 evaluated LVLMs struggle with structured reasoning and generalization, often failing to link abstract knowledge with anatomically grounded visual interpretation. The code is available at https://github.com/ttumyche/CXReasonBench
The quadratic complexity of the attention mechanism severely limits the context scalability of Video Diffusion Transformers (DiTs). We find that the highly sparse spatio-temporal attention patterns exhibited in Video DiTs can be naturally represented by the Monarch matrix. It is a class of structured matrices with flexible sparsity, enabling sub-quadratic attention via an alternating minimization algorithm. Accordingly, we propose VMonarch, a novel attention mechanism for Video DiTs that enables efficient computation over the dynamic sparse patterns with structured Monarch matrices. First, we adapt spatio-temporal Monarch factorization to explicitly capture the intra-frame and inter-frame correlations of the video data. Second, we introduce a recomputation strategy to mitigate artifacts arising from instabilities during alternating minimization of Monarch matrices. Third, we propose a novel online entropy algorithm fused into FlashAttention, enabling fast Monarch matrix updates for long sequences. Extensive experiments demonstrate that VMonarch achieves comparable or superior generation quality to full attention on VBench after minimal tuning. It overcomes the attention bottleneck in Video DiTs, reduces attention FLOPs by a factor of 17.5, and achieves a speedup of over 5x in attention computation for long videos, surpassing state-of-the-art sparse attention methods at 90% sparsity. Code is available at https://github.com/EricLina/VMonarch.
시계열 예측은 주기 내 변동과 주기 간 트렌드의 복잡한 얽힘으로 인해 여전히 어려운 문제로 남아 있다. 최근 연구들이 1D 시퀀스를 2D 주기-위상 표현으로 변환하려는 시도를 해왔으나, 두 가지 주요 한계를 지닌다. 첫째, 변환된 텐서를 정적 이미지로 처리하면 위상 불일치가 발생하는데, 표준 공간 연산자가 그리드 경계에서 시간적 연속성을 단절시키기 때문이다. 둘째, 균일한 고정 크기 표현에 의존하면 모델링 용량이 비효율적으로 배분되며, 압축 가능한 비정상(non-stationary) 시간 패턴에 필요한 적응적 해상도를 제공하지 못한다. 이러한 한계를 해결하기 위해, 우리는 예측 패러다임을 회귀에서 2D 생성적 렌더링으로 근본적으로 전환하는 새로운 프레임워크 TimeGS를 제안한다. 미래 시퀀스를 잠재적 2D 시간 표면으로 재개념화함으로써, TimeGS는 가우시안 커널의 고유한 비등방성(anisotropy)을 활용하여 유연한 기하학적 정렬로 복잡한 변동을 적응적으로 모델링한다. 이를 실현하기 위해, 최적화를 안정화하고자 고정 사전(dictionary)에서 커널을 합성하는 Multi-Basis Gaussian Kernel Generation(MB-GKG) 블록과, 주기적 경계에 걸친 엄격한 시간적 연속성을 강제하는 Multi-Period Chronologically Continuous Rasterization(MP-CCR) 블록을 도입한다. 표준 벤치마크 데이터셋에 대한 종합적인 실험을 통해 TimeGS가 최첨단 또는 그에 준하는 성능을 달성함을 입증한다. 코드는 https://github.com/yixinwang1/TimeGS 에서 확인할 수 있다.
검색-증강 생성(RAG)은 대규모의 이질적인 문서 컬렉션으로 규모를 확장할 때 성능이 저하되는데, 이때 밀도 유사도는 식별력을 잃고 top-k 검색은 의미적으로 유사하지만 문맥적으로 부정확한 청크를 점점 더 많이 반환합니다. 우리는 이러한 실패 모드를 벡터 검색 희석(vector search dilution)이라고 부릅니다. 하이브리드 밀도+희소 검색을 사용하더라도, 우리는 배포된 와이오밍 교통부(Wyoming Department of Transportation) 말뭉치에서 이를 직접 목격했는데, 54개 문서에서 1,128개 문서(88,907개 청크)로 규모를 확장하니 정확도가 75%에서 40% 미만으로 떨어졌습니다. 이러한 희석을 해결하기 위해 우리는 MASDR-RAG(RAG를 위한 다중 에이전트 범위 지정 도메인 검색)를 제안하고, 5개의 LLM 백본, 6개의 말뭉치, 2개의 인덱스 스택에 걸쳐 200개의 전문가 검증 쿼리를 사용하여 평가했습니다. 우리의 결과는 조직 메타데이터를 사용한 도메인 범위 지정이 핵심 해결책임을 시사하며, P@10을 0.77에서 0.86으로 유의미하게 향상시켰습니다($p < 0.05$). 또한, 다중 에이전트 오케스트레이션에 대한 조사를 통해 높은 수준의 구성 의존성이 발생함을 밝혔는데, 이는 우리가 정밀도-충실도 역설(precision-faithfulness paradox)이라고 부르는 현상을 만듭니다. 이러한 다양한 결과를 바탕으로, 우리의 실질적인 권장 사항은 간단합니다. 먼저 범위를 지정한 다음 단일 합성 호출을 수행하고, 완전한 다중 에이전트 오케스트레이션은 네이티브 툴 호출 백본과 결합된 진정한 다중 도메인 말뭉치를 위해 남겨두십시오. 코드와 데이터는 수락 시 공개될 예정입니다.
확률적 편미분방정식(SPDE) 대리 모델(surrogate)을 평가하기 위한 프레임워크 TRIE를 제안하며, 불변 측도 재현성, 예측 불확실성의 신뢰성, 효율적 확률적 생성 능력을 기준으로 평가한다. stochastic Kuramoto–Sivashinsky와 stochastic Kolmogorov flow에 대한 실험 결과, 결정론적 대리 모델은 단기 예측에는 유효하나 장기 통계 구조 재현에 실패하며, 생성 모델이 가장 일관된 성능을 보이고 잠재 생성 모델은 추론 시간을 약 12배 단축하면서도 통계적 충실도를 유지함을 확인했다.
AI 에이전트는 이제 다양한 하드웨어 플랫폼에서 저수준 연산자(operator) 커널을 작성하고, 컴파일하며, 반복적으로 최적화할 수 있다. 그러나 기존 벤치마크는 거의 전적으로 CUDA와 Triton에 초점을 맞추어, 프로그래밍 모델 노출이 적은 하드웨어 생태계에는 공통 평가 기준선이 없다. 우리는 Huawei Ascend NPU에서 AI 생성 연산자 코드를 위한 공개 벤치마크인 CANN Bench를 제안한다. 현재 릴리스는 단순 elementwise 기본 연산부터 MoE dispatch 및 FlashAttention 커널까지, FP16, BF16, FP32, INT8 정밀도 형식을 아우르는 53개 연산자와 1060개 테스트 케이스를 네 가지 난이도 등급으로 구성한다. 평가는 컴파일, 기능적 정확성, 성능을 독립 축으로 취급하는 3차원 가중 복합 점수를 채택하여, 커널 생성 에이전트에 원칙적인 보상 신호를 제공한다. 성능은 즉시 사용 가능한 PyTorch-on-Ascend 기준선과 실제 NPU 하드웨어에서의 케이스별 분석적 Hardware-Anchored Performance(HAP) 한계에 대해 채점되어, 점수가 측정 아티팩트가 아닌 진정한 최적화 여지를 반영하도록 한다. 평가 하네스는 근본적으로 보상 해킹(reward hacking)에 저항하도록 설계되었다. CANN Bench는 공식 CANN 저장소 내에서 버전 관리되며, 장기적인 커뮤니티 공동 구축을 위해 설계되어 Ascend 생태계에 AI 연산자 작성 역량에 대한 정량적이고 재현 가능하며 지속적으로 유지되는 척도를 제공한다.
이론 연구의 자동화는 후보 결과의 생성뿐만 아니라 이에 대한 신뢰성 있는 평가에 의해서도 제약을 받습니다. 일반적인 접근 방식은 대규모 언어 모델(LLM) 리뷰어를 통해 연구 루프를 완결하는 것입니다. 그러나 이러한 리뷰어는 실증적으로 여전히 신뢰하기 어렵습니다. 위조된 논문을 수락할 수 있으며, 이를 무작위 확률에 가까운 비율로만 탐지하기도 합니다 (Bad Scientist, 2025). 본 논문에서는 Lean 증명 보조기(proof assistant)에 기반하여 인과 추론(causal inference) 분야의 이론 연구를 자동화하는 프레임워크인 CausalSmith를 제시합니다. CausalSmith는 인간의 설계 및 검토 하에 언어 모델의 지원을 받아 개발된 7,035개의 기계 검증 선언을 포함하는 인과 추론 기초 Lean 라이브러리인 Causalean과, 연구 주제 선정, 결과 제안, 진술 정형화, 증명 구성, 인간 검사를 위한 최종 산출물 제시를 수행하는 자기 개선 에이전트 파이프라인을 결합합니다. 기계 검증 증명은 정형 명제가 그 가정으로부터 도출된다는 점만을 확립할 뿐 해당 명제가 의도한 과학적 주장을 충실히 담아내는지까지는 보장하지 못하므로, 본 파이프라인은 각 정형 정리를 표현하고자 했던 비정형 주장과 비교하는 진술 감사(statement audit)로 커널(kernel) 검증을 보강합니다. 우리는 완료된 자율 연구 실행을 통해 생성된 산출물을 사용하여 시스템을 평가합니다. 소스 코드, 정형 라이브러리 및 실행 기록은 https://github.com/Jiyuan-Tan/CausalSmith 에서 확인할 수 있습니다.
다중모달 대규모 언어 모델(MLLM)에서의 추론은 세밀한 시각적 인지(perception)와 엄밀한 논리적 연역을 모두 필요로 합니다. 텍스트 기반의 명시적 사고의 연쇄(Chain-of-Thought, CoT)는 계산 비용이 많이 들고 시각적 환각(hallucination)에 취약한 반면, 기존의 잠재적 추론(latent reasoning) 방식은 일반적으로 많은 비용이 드는 학습을 요구합니다. 나아가 학습이 불필요한 LLM 추론 메커니즘을 다중모달 환경에 직접 적용하면 불안정한 성능을 초래합니다. 본 연구에서는 이러한 실패가 토큰 수준의 엔트로피에 의존하기 때문에 발생하며, 이는 본질적으로 인지적 모호성(예: 불명확한 시각적 세부 정보)과 논리적 불확실성(예: 복잡한 추론 단계)을 혼동한다는 점을 밝혀냈습니다. 이러한 병목 현상을 극복하기 위해, 인지와 추론을 명시적으로 분리하는 MLLM용 새로운 학습 불필요 추론 전략을 제시합니다. 모델의 인지적 초점을 동적으로 측정하기 위한 새로운 척도인 시각-텍스트 어텐션 비율(vision-to-text attention ratio)을 제안합니다. 이 척도를 바탕으로, 제안하는 프레임워크인 AGS(Attention-Guided Switching)는 연속 공간에서 고충실도 시각 정보를 보존하기 위해 인지 토큰에 대해서는 잠재적 추론을 적응적으로 유도하고, 구조적 닻내림(anchoring)을 유지하기 위해 논리 토큰에 대해서는 명시적인 텍스트 생성을 강제합니다. 광범위한 실험을 통해 제안 방법이 최고 수준의 성능을 달성하며, 자기회귀 단계와 지연 시간을 줄여 정확도와 추론 효율성을 모두 크게 향상시킴을 입증합니다. 코드는 https://github.com/swordAndSnow/MM26-AGS 에 공개되어 있습니다.
본 프리프린트는 연구 지향적이고 체계적인 실습 과정을 제시하며, 토큰화(tokenisation)와 벡터화(vectorisation)부터 대규모 언어 모델(LLM) 파인튜닝(fine-tuning), 검색-증강 생성(RAG), 인간 피드백 강화학습(RLHF), 프롬프트 엔지니어링, 모델 압축, 멀티모달 시스템에 이르는 현대 NLP 파이프라인 전반을 독자에게 안내한다. 17개의 실습 세션은 간결한 이론과 상세한 구현 계획, 형식화된 평가 지표, 투명한 평가 기준을 결합한다. 이 작업은 기존 교과서가 아니며, 각 세션마다 공개 저장소에 코드, 모델, 보고서를 게시해야 하는 재현 가능한 연구 산출물로 설계되었다. 모든 실험은 하나의 점진적으로 확장되는 단일 말뭉치(corpus) 위에서 수행되며, 상업용 API보다 오픈 가중치(open weight) 모델을 권장하고 Hugging Face 생태계에 특별히 주목한다. 자료는 저자원 언어에 대한 독자적 연구로 보강되며, 타지크어와 타타르어를 위한 언어 자원—서브워드 토크나이저, 임베딩(embedding), 어휘 데이터베이스, 음역(transliteration) 벤치마크—을 포함하여, 현대 NLP가 데이터가 부족한 환경에 어떻게 적응할 수 있는지 보여준다. 실습 과정은 AI 에이전트, 멀티 에이전트 시스템, LLMOps, 효율적 추론 등 고급 주제도 다루며, 연구와 산업 배포 모두에 대비할 수 있도록 한다. 고학년 학부생, 대학원생, 그리고 고전적 ML부터 최신 멀티모달 및 에이전트 기반 시스템에 이르는 방법을 구현·비교·배포하려는 실무 개발자를 대상으로 설계되었다.