Tran Son

NeurIPS Conference 2025 Conference Paper

Directed-Tokens: A Robust Multi-Modality Alignment Approach to Large Language-Vision Models

• Thanh-Dat Truong
• Huu-Thien Tran
• Tran Son
• Bhiksha Raj
• Khoa Luu

Large multimodal models (LMMs) have gained impressive performance due to their outstanding capability in various understanding tasks. However, these models still suffer from some fundamental limitations related to robustness and generalization due to the alignment and correlation between visual and textual features. In this paper, we introduce a simple but efficient learning mechanism for improving the robust alignment between visual and textual modalities by solving shuffling problems. In particular, the proposed approach can improve reasoning capability, visual understanding, and cross-modality alignment by introducing two new tasks: reconstructing the image order and the text order into the LMM's pre-training and fine-tuning phases. In addition, we propose a new directed-token approach to capture visual and textual knowledge, enabling the capability to reconstruct the correct order of visual inputs. Then, we introduce a new Image-to-Response Guided loss to further improve the visual understanding of the LMM in its responses. The proposed approach consistently achieves state-of-the-art (SoTA) performance compared with prior LMMs on academic task-oriented and instruction-following LMM benchmarks.

AAAI Conference 2016 Conference Paper

Solving Goal Recognition Design Using ASP

• Tran Son
• Orkunt Sabuncu
• Christian Schulz-Hanke
• Torsten Schaub
• William Yeoh

Goal Recognition Design involves identifying the best ways to modify an underlying environment that agents operate in, typically by making a subset of feasible actions infeasible, so that agents are forced to reveal their goals as early as possible. Thus far, existing work has focused exclusively on imperative classical planning. In this paper, we address the same problem with a different paradigm, namely, declarative approaches based on Answer Set Programming (ASP). Our experimental results show that one of our ASP encodings is more scalable and is significantly faster by up to three orders of magnitude than the current state of the art.

AAAI Conference 2015 Conference Paper

Exploring the KD45 Property of a Kripke Model After the Execution of an Action Sequence

• Tran Son
• Enrico Pontelli
• Chitta Baral
• Gregory Gelfond

The paper proposes a condition for preserving the KD45n property of a Kripke model when a sequence of update models is applied to it. The paper defines the notions of a primitive update model and a semi-reflexive KD45n (or sr-KD45n) Kripke model. It proves that updating a sr-KD45n Kripke model using a primitive update model results in a sr-KD45n Kripke model, i. e. , a primitive update model preserves the properties of a sr-KD45n Kripke model. It shows that several update models for modeling well-known actions found in the literature are primitive. This result provides guarantees that can be useful in presence of multiple applications of actions in multi-agent system (e. g. , multi-agent planning).

AAAI Conference 2015 Conference Paper

Solving Distributed Constraint Optimization Problems Using Logic Programming

• Tiep Le
• Tran Son
• Enrico Pontelli
• William Yeoh

This paper explores the use of answer set programming (ASP) in solving distributed constraint optimization problems (DCOPs). It makes the following contributions: (i) It shows how one can formulate DCOPs as logic programs; (ii) It introduces ASP-DPOP, the first DCOP algorithm that is based on logic programming; (iii) It experimentally shows that ASP-DPOP can be up to two orders of magnitude faster than DPOP (its imperative-programming counterpart) as well as solve some problems that DPOP fails to solve due to memory limitations; and (iv) It demonstrates the applicability of ASP in the wide array of multi-agent problems currently modeled as DCOPs.

AAAI Conference 2011 Conference Paper

Conjunctive Representations in Contingent Planning: Prime Implicates Versus Minimal CNF Formula

• Son To
• Tran Son
• Enrico Pontelli

This paper compares in depth the effectiveness of two conjunctive belief state representations in contingent planning: prime implicates and minimal CNF, a compact form of CNF formulae, which were initially proposed in conformant planning research (To et al. 2010a; 2010b). Similar to the development of the contingent planner CNFct for minimal CNF (To et al. 2011b), the present paper extends the progression function for the prime implicate representation in (To et al. 2010b) for computing successor belief states in the presence of incomplete information to handle non-deterministic and sensing actions required in contingent planning. The idea was instantiated in a new contingent planner, called PIct, using the same AND/OR search algorithm and heuristic function as those for CNFct. The experiments show that, like CNFct, PIct performs very well in a wide range of benchmarks. The study investigates the advantages and disadvantages of the two planners and identifies the properties of each representation method that affect the performance.

AAAI Conference 2011 Conference Paper

On Improving Conformant Planners by Analyzing Domain-Structures

• Khoi Nguyen
• Vien Tran
• Tran Son
• Enrico Pontelli

The paper introduces a novel technique for improving the performance and scalability of best-first progression-based conformant planners. The technique is inspired by different wellknown techniques from classical planning, such as landmark and stratification. Its most salient feature is that it is relatively cheap to implement yet quite effective when applicable. The effectiveness of the proposed technique is demonstrated by the development of new conformant planners by integrating the technique in various state-of-the-art conformant planners and an extensive experimental evaluation of the new planners using benchmarks collected from various sources. The result shows that the technique can be applied in several benchmarks and helps improve both performance and scalability of conformant planners.

AAAI Conference 2011 Conference Paper

On the Effectiveness of Belief State Representation in Contingent Planning

• Son To
• Tran Son
• Enrico Pontelli

AAAI Conference 2010 Conference Paper

On the Use of Prime Implicates in Conformant Planning

• Son To
• Tran Son
• Enrico Pontelli

The paper presents an investigation of the use of two alternative forms of CNF formulae—prime implicates and minimal CNF—to compactly represent belief states in the context of conformant planning. For each representation, we define a transition function for computing the successor belief state resulting from the execution of an action in a belief state; results concerning soundness and completeness are provided. The paper describes a system (PIP) which dynamically selects either of these two forms to represent belief states, and an experimental evaluation of PIP against state-of-the-art conformant planners. The results show that PIP has the potential of scaling up better than other planners in problems rich in disjunctive information about the initial state.