Papers by Nguyen Cam-Tu
LLMs Meet Isolation Kernel: Lightweight, Learning-free Binary Embeddings for Fast Retrieval (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) embeddings are typically high-dimensional, leading to substantial storage and retrieval overhead. |
| Approach: | They propose a learning-free method that transforms an LLM embedding into a binary embeddable using Isolation Kernel (IKE). |
| Outcome: | The proposed method performs 16.7 faster retrieval and 16 lower memory usage than the original LLM embeddings while maintaining comparable accuracy. |
Reward Difference Optimization For Sample Reweighting In Offline RLHF (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are becoming more capable, but their maximum likelihood objective for the next token prediction falls short in capturing such crucial human values. |
| Approach: | They propose a reward difference prediction method that uses reward difference coefficients to reweigh sample pairs in offline RLHF and a difference model that considers rich interactions between a pair of responses. |
| Outcome: | The proposed method is effective in both automatic metrics and human evaluation. |
Improving Question Generation with Multi-level Content Planning (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies suggest key phrase selection is essential for question generation, yet it is difficult to connect disjointed phrases into meaningful questions, especially for long context. |
| Approach: | They propose a QG framework that uses multi-level content planning to generate questions from a given context and an answer. |
| Outcome: | The proposed framework outperforms baselines on two popular QG datasets. |
Consultant Decoding: Yet Another Synergistic Mechanism (2025.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) have attracted widespread attention and adoption across diverse domains due to their exceptional performance and robust generalization abilities. |
| Approach: | They propose a synergetic mechanism for Consultant Decoding (CD) that achieves a 2.5-fold increase in inference speed compared to the target model while maintaining comparable generation quality. |
| Outcome: | The proposed mechanism achieves 2.5-fold increase in inference speed while maintaining comparable generation quality (100% of the target model’s performance). |
Retrospex: Language Agent Meets Offline Reinforcement Learning Critic (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing LLM agent frameworks do not fully utilize past experiences for improvement. |
| Approach: | They propose a LLM-based agent framework called Retrospex that analyzes past experiences in depth to improve existing agent frameworks. |
| Outcome: | The proposed framework analyzes past experiences in ScienceWorld, ALFWorld and Webshop environments, demonstrating its advantages over baselines. |
TrigReason: Trigger-Based Collaboration between Small and Large Reasoning Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Reasoning Models suffer from high inference latency due to autoregressive reasoning . SpecReason adopts a polling-based design that repeatedly invokes the LRM for verification at every step . |
| Approach: | They propose a trigger-based collaborative reasoning framework that delegates most reasoning to the SRM and activates LRM intervention only when necessary. |
| Outcome: | The proposed framework reduces latency and API cost by 73.3% under edge–cloud conditions. |
Diversify Question Generation with Retrieval-Augmented Style Transfer (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing question generation systems focus on the internal knowledge within the textual passage or the semantic word space for diverse content planning. Existing solutions focus on relying on the knowledge of the text and the semantic words, but have not considered the potential of external knowledge for expression diversity. |
| Approach: | They propose a framework for Retrieval-Augmented Style Transfer that utilizes the style of diverse templates for question generation. |
| Outcome: | The proposed framework outperforms baselines on diversity while being comparable in terms of consistency scores. |
daDPO: Distribution-Aware DPO for Distilling Conversational Abilities (2025.findings-acl)
Copied to clipboard
| Challenge: | Knowledge distillation (KD) with Direct Preference Optimization (DPO) has emerged as a promising approach to enhance the conversational abilities of smaller models using a larger teacher model. |
| Approach: | They propose a framework that integrates the teacher's distributional information into DPO distillation while preserving theoretical guarantees. |
| Outcome: | The proposed framework outperforms existing methods in restoring performance for pruned models and enhancing smaller models within the same LLM family. |
MiMIC: Mitigating Visual Modality Collapse in Universal Multimodal Retrieval While Avoiding Semantic Misalignment (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing UMR methods can be broadly divided into two categories: early-fusion approaches, such as Marvel, which projects visual features into the language model space for integrating with text modality, and late-fusion methods, such UniVL-DR, which encode visual and textual inputs using separate encoders and obtain fused embeddings through addition. |
| Approach: | They propose to map different modalities into a shared embedding space for multi-modal retrieval. |
| Outcome: | Experiments on the WebQA+ and EVQA+ datasets show that MiMIC outperforms both early- and late-fusion approaches. |
Thermometer of Thoughts: Enhancing LLM’s Exploration via Attention Temperature Modulation (2026.acl-long)
Copied to clipboard
| Challenge: | Recent advances in the reasoning capabilities of large language models have enabled them to tackle complex tasks such as mathematics reasoning. |
| Approach: | They propose a method that modulates attention temperature dynamically to shift LLM's internal focus during reasoning, enabling a dynamic shift between exploratory and focused modes. |
| Outcome: | The proposed method improves Pass@10 by 6.78–14.20% and aggregation accuracy by 9.74% across 7 reasoning benchmarks. |
MLAlgo-Bench: Can Machines Implement Machine Learning Algorithms? (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Currently, the top-performing models achieve a 48.8% task completion rate on realizing machine learning algorithms . |
| Approach: | They propose a benchmark to test machine learning's ability to generate ML code for humans . they propose an automatic evaluation framework with metrics such as task pass rate and time overhead . |
| Outcome: | The proposed benchmark is unique in its focus on interpreting complex human instructions and producing multi-step, high-complexity code. |
Bidirectional Semantic Enhancement for Schema Routing Across Large-Scale Databases (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods relying on unidirectional query expansion fail to bridge lexical mismatches and graph-based approaches struggle to navigate schemas when explicit structural links are missing. |
| Approach: | They propose a retrieval framework that bridges the semantic gap between user queries and vague schema definitions by performing online generative query expansion. |
| Outcome: | The proposed retrieval framework bridges the semantic gap between user queries and vague schema definitions by enriching table schemas offline and performing online generative query expansion. |
LAVa: Layer-wise KV Cache Eviction with Dynamic Budget Allocation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for cache compression are heuristic and lack dynamic budget allocation . cnn's john mccartney and johnny mccain present a new approach for cache eviction and dynamic budgets . |
| Approach: | They propose a unified framework for cache compression that minimizes information loss in transformer residual streams. |
| Outcome: | The proposed method consistently maintains top performance across task types. |