Research

Published work across language models, human-centered NLP, and efficient computing.

Agentic RAG, Tool Use & Retrieval Systems

Reinforcement learning · Tool use

GRASP: Granularity-Aware Search Policy for Agentic RAG

Learning when to search, read, and answer. GRASP coordinates semantic search, keyword search, and paragraph reading to acquire evidence for multi-hop questions.

Findings of EMNLP 2026

Read paper Project & code

Post-Training, Dialogue Systems & Model Behavior

User simulation · Model evaluation

Simulated Students in Tutoring Dialogues: Substance or Illusion?

Evaluating whether simulated students reproduce the linguistic, behavioral, and cognitive properties of real learners in tutoring dialogues.

ACL 2026

Read paper

Preference optimization · Activation steering

Letting Tutor Personas Speak Up for LLMs: Learning Steering Vectors from Dialogue via Preference Optimization

Learning tutor-persona steering vectors from human tutor–student dialogues to control model behavior through observed conversational preferences.

BEA 2026

Read paper

Post-training · Tutoring dialogues

Training LLM-based Tutors to Improve Student Learning Outcomes in Dialogues

Training language-model tutors with the goal of improving student learning across a dialogue.

AIED 2025

Read paper

Applied NLP, Personalization & Human-Centered Evaluation

Personalization · Human–AI interaction

Personalizing Kanji Memorization: Designing Adaptive Mnemonics Based on Learner Preferences Using Large Language Models

Designing adaptive kanji mnemonics around learner preferences through an interactive prototype and user feedback.

CHI EA 2026

Read paper

Preference modeling · Personalized generation

Interpretable Mnemonic Generation for Kanji Learning via Expectation-Maximization

Inferring community-level and user-specific preferences to personalize mnemonic generation with limited feedback.

EMNLP 2025

Read paper

Cross-lingual NLP · Mnemonic generation

PhoniTale: Phonologically Grounded Mnemonic Generation for Typologically Distant Language Pairs

Using phonological similarity across languages to find keywords for memorable vocabulary-learning cues.

EMNLP 2025

Read paper

Applied generation · Educational content

From Text to Visuals: Using LLMs to Generate Math Diagrams with Vector Graphics

Generating vector-based mathematical diagrams from text with large language models.

AIED 2025

Read paper

Mnemonic generation · Human evaluation

Exploring Automated Keyword Mnemonics Generation with Large Language Models via Overgenerate-and-Rank

Generating and ranking candidate mnemonics, with evaluation of their quality and usefulness as learning aids.

Findings of EMNLP 2024

Read paper

Assessment generation · Applied NLP

Exploring Automated Distractor Generation for Math Multiple-choice Questions via Large Language Models

Using large language models to generate distractors for mathematics multiple-choice questions.

Findings of NAACL 2024

Read paper

Language learning · Multimodal mnemonics

SmartPhone: Exploring Keyword Mnemonic with Auto-generated Verbal and Visual Cues

Exploring automatically generated verbal and visual cues for keyword-based vocabulary learning.

AIED 2023

Read paper

ML Systems, Compilers & Efficient Inference

Model partitioning · Processing-in-memory

Optimal Model Partitioning with Low-Overhead Profiling on the PIM-based Platform for Deep Learning Inference

Partitioning deep-learning inference across heterogeneous hardware using low-overhead profiling and runtime scheduling.

ACM TODAES

Read paper

ML infrastructure · Runtime integration

Bringing PIM into ONNX Runtime

Connecting ONNX models to processing-in-memory operations and evaluating BERT inference on an FPGA-emulated PIM platform.

ICEIC 2022

Read paper

Compilers · Memory management

Tile-based Code Generation for Scratchpad Memory

Extending LLVM/Polly to generate tiled execution, memory transfers, and address calculations for scratchpad memory.

ICEIC 2021

Read paper

Research

Natural language processing and ML systems, from model behavior to efficient inference.

01 / Natural language processing

Understanding and shaping model behavior.

All publications on Scholar

NLP × Systems · Agentic retrieval

Learning when to search,
read, and answer.

GRASP trains language-model agents to coordinate semantic search, keyword search, and paragraph reading. The policy learns how to acquire evidence and when to expand its context.

Explore GRASP Project & code

01Search broadly
Find clues with semantic retrieval

02Read selectively
Expand a sentence’s parent paragraph

03Search precisely
Resolve entities with keyword retrieval

An illustration of the available actions. The learned policy chooses their order and when to answer.

Evaluation · User simulation

Beyond a convincing conversation.

When do LLM-simulated students behave like real learners? We evaluate linguistic, behavioral, and cognitive realism, showing why surface-level plausibility can hide failures in learning dynamics.

Substance or Illusion?ACL 2026

Post-training · Activation steering

Learning behavior from real dialogue.

Tutor-persona steering vectors learned from human tutor–student dialogues offer interpretable control over model behavior. Preference optimization connects observed conversations to how a model responds.

Letting Tutor Personas Speak UpBEA 2026 · First author

Personalization · Human-centered AI

A memory aid that fits the learner.

From interpretable preference modeling to an interactive iOS prototype, this work explores how language models can generate kanji mnemonics that adapt to individual learners, even with limited feedback.

EM personalization Adaptive mnemonic designEMNLP 2025 · CHI EA 2026

A closer look / GRASP

What should an agent read next?

Multi-hop questions require more than retrieving a relevant document. An agent must identify the next missing piece of evidence, choose a useful tool, and decide how much context to expose.

GRASP uses GRPO to learn these decisions. Search returns sentences; a separate read action expands their parent paragraphs. The training signal combines answer accuracy, grounded reading, complementary search, and turn efficiency.

The updated study evaluates both 3B and 7B models on HotpotQA, 2WikiMultihopQA, and MuSiQue.

Explore the project

72 F1

HotpotQA · 7B model · 500 sampled validation questions

GRASP: 72

AutoRefine: 65

Token F1 × 100. Strongest listed baseline on this metric: AutoRefine. Updated manuscript, Table 3. GRASP ties AutoRefine on MuSiQue F1.

What remains open

The training rewards rely on gold supporting-document annotations. Learning grounded tool use with weaker supervision is an important next step.

02 / ML Systems

Making inference work efficiently.

The model is one part of the system. Retrieval, data movement, and where computation runs shape its practical performance.

Hardware–software co-design · Inference

Choosing where a model should run.

Model partitioning and low-overhead profiling for deep-learning inference on a processing-in-memory platform. This work connects runtime decisions to the capabilities of heterogeneous hardware.

Model partitioningACM TODAES · Co-first author

ML infrastructure · Runtime integration

Bringing PIM into ONNX Runtime.

An execution provider connects ONNX models to processing-in-memory operations. The framework offloads supported operators and evaluates end-to-end BERT inference on an FPGA-emulated PIM platform.

ONNX Runtime paperICEIC 2022

Compilers · Memory management

Generating code around the memory hierarchy.

Extending LLVM/Polly to generate tiled execution, memory transfers, and address calculations for scratchpad memory. The study examines how tile shapes and reuse affect data movement.

Tile-based code generationICEIC 2021 · First author

Connecting NLP and systems

GRASP studies retrieval decisions through learned policies. My ongoing RAG systems work also examines the latency–accuracy trade-offs of approximate versus exact retrieval, including time-to-first-token and downstream answer quality.

Current research in the CV

Agentic RAG, Tool Use & Retrieval Systems

GRASP: Granularity-Aware Search Policy for Agentic RAG

Post-Training, Dialogue Systems & Model Behavior

Simulated Students in Tutoring Dialogues: Substance or Illusion?

Letting Tutor Personas Speak Up for LLMs: Learning Steering Vectors from Dialogue via Preference Optimization

Training LLM-based Tutors to Improve Student Learning Outcomes in Dialogues

Applied NLP, Personalization & Human-Centered Evaluation

Personalizing Kanji Memorization: Designing Adaptive Mnemonics Based on Learner Preferences Using Large Language Models

Interpretable Mnemonic Generation for Kanji Learning via Expectation-Maximization

PhoniTale: Phonologically Grounded Mnemonic Generation for Typologically Distant Language Pairs

From Text to Visuals: Using LLMs to Generate Math Diagrams with Vector Graphics

Exploring Automated Keyword Mnemonics Generation with Large Language Models via Overgenerate-and-Rank

Exploring Automated Distractor Generation for Math Multiple-choice Questions via Large Language Models

SmartPhone: Exploring Keyword Mnemonic with Auto-generated Verbal and Visual Cues

ML Systems, Compilers & Efficient Inference

Optimal Model Partitioning with Low-Overhead Profiling on the PIM-based Platform for Deep Learning Inference

Bringing PIM into ONNX Runtime

Tile-based Code Generation for Scratchpad Memory

Understanding and shaping model behavior.

Learning when to search,read, and answer.

Beyond a convincing conversation.

Learning behavior from real dialogue.

A memory aid that fits the learner.

What should an agent read next?

Making inference work efficiently.

Choosing where a model should run.

Bringing PIM into ONNX Runtime.

Generating code around the memory hierarchy.

Learning when to search,
read, and answer.