Selected work

Representative Papers

AdMem: Advanced Memory for Task-solving Agents

Jun 04, 2026

This work addresses the challenges large language models face in long-horizon tasks regarding knowledge retention, organization, and reuse. Existing approaches are often limited to static fact storage or replay of successful experiences, struggling to incorporate failure cases and lacking online extensibility. To overcome these limitations, the paper proposes a unified memory framework that, for the first time, integrates semantic, episodic, and procedural memory within a dual-layer short- and long-term storage architecture. A multi-agent design—comprising actor, memory, and critic components—enables automatic memory generation, reward-based annotation, and adaptive retrieval. A reward-driven long-term memory management strategy, including evaluation, consolidation, and pruning, facilitates continual learning and online evolution. Experiments demonstrate that the proposed method significantly improves success rates and robustness across diverse complex long-horizon tasks, outperforming current baselines.

0 citationsRead paper

Effective and Memory-Efficient Alternatives to ECC for Reliable Large-Scale DNNs

May 08, 2026

This work addresses the vulnerability of deep neural networks to memory transient faults in safety-critical applications, where conventional error-correcting codes like SECDED incur substantial hardware overhead. The authors propose two lightweight protection mechanisms: MSET selectively hardens the most critical parameter bits in CNNs and Vision Transformers (ViTs) through vulnerability analysis, while CEP offers fine-grained, full-parameter protection. Notably, they demonstrate for the first time that safeguarding only the most significant exponent bit of FP16/FP32 floating-point numbers significantly enhances ViT reliability. Experimental results show that the proposed approaches enable up to 10× higher tolerance to bit error rates on large-scale CNNs and ViTs, reduce area overhead by 3.5×, accelerate decoding by 7×, and achieve superior reliability compared to SECDED—all without additional storage cost.

0 citationsRead paper

SPEC CPU: The Next Generation

May 02, 2026

This work addresses the limitations of existing CPU benchmarks in accurately evaluating the performance of modern heterogeneous, multithreaded processors under diverse workloads. To this end, the authors present the SPEC CPU 2026 benchmark suite, developed through community collaboration and principled methodology, which introduces the Rolling-Round-Robin Rate approach to standardize the execution of heterogeneous multiprogrammed workloads. The suite incorporates newly designed multithreaded benchmarks exhibiting varied microarchitectural characteristics, selected and hardened through an open-source application curation process. Emphasizing workload diversity, portability, and long-term viability, SPEC CPU 2026 establishes a robust, representative, and authoritative standard for performance evaluation, thereby supporting next-generation computer architecture research.

0 citationsRead paper

I hope we don't do to trust what advertising has done to love

Apr 30, 2026

This work addresses the prevalent misuse and conceptual vagueness surrounding “trust” in contemporary artificial intelligence, particularly within agent-based systems—a phenomenon akin to the commodification of “love” in advertising. The paper introduces, for the first time, a structured “pillars of trust” framework that operationalizes abstract notions of trust into measurable, actionable dimensions. Through human-AI interaction design, these dimensions are concretized as a “trust vector.” This framework not only facilitates interdisciplinary dialogue across computational domains and civil society but also establishes a theoretical foundation and public deliberation mechanism for the design and evaluation of trustworthy AI systems, thereby advancing the standardization and transparency of trust in technical practice.

0 citationsRead paper

Crystal structure prediction using graph neural combinatorial optimization

Apr 26, 2026

This work addresses the discrete combinatorial optimization challenge in crystal structure prediction arising from the absence of symmetry constraints by introducing, for the first time, graph neural combinatorial optimization to this domain. The proposed method constructs an extended graph to simultaneously model both short-range and long-range atomic interactions and integrates the Gumbel-Sinkhorn mechanism to efficiently generate crystal structures that satisfy a target stoichiometry within an unsupervised framework. Evaluated across diverse chemical compositions, the approach significantly outperforms classical heuristic algorithms and achieves performance comparable to commercial optimization solvers, thereby establishing a new paradigm for GPU-accelerated, large-scale crystal structure prediction.

0 citationsRead paper

Recent publications

Latest Papers

AdMem: Advanced Memory for Task-solving Agents

Jun 04, 2026

This work addresses the challenges large language models face in long-horizon tasks regarding knowledge retention, organization, and reuse. Existing approaches are often limited to static fact storage or replay of successful experiences, struggling to incorporate failure cases and lacking online extensibility. To overcome these limitations, the paper proposes a unified memory framework that, for the first time, integrates semantic, episodic, and procedural memory within a dual-layer short- and long-term storage architecture. A multi-agent design—comprising actor, memory, and critic components—enables automatic memory generation, reward-based annotation, and adaptive retrieval. A reward-driven long-term memory management strategy, including evaluation, consolidation, and pruning, facilitates continual learning and online evolution. Experiments demonstrate that the proposed method significantly improves success rates and robustness across diverse complex long-horizon tasks, outperforming current baselines.

0 citationsRead paper

Effective and Memory-Efficient Alternatives to ECC for Reliable Large-Scale DNNs

May 08, 2026

This work addresses the vulnerability of deep neural networks to memory transient faults in safety-critical applications, where conventional error-correcting codes like SECDED incur substantial hardware overhead. The authors propose two lightweight protection mechanisms: MSET selectively hardens the most critical parameter bits in CNNs and Vision Transformers (ViTs) through vulnerability analysis, while CEP offers fine-grained, full-parameter protection. Notably, they demonstrate for the first time that safeguarding only the most significant exponent bit of FP16/FP32 floating-point numbers significantly enhances ViT reliability. Experimental results show that the proposed approaches enable up to 10× higher tolerance to bit error rates on large-scale CNNs and ViTs, reduce area overhead by 3.5×, accelerate decoding by 7×, and achieve superior reliability compared to SECDED—all without additional storage cost.

0 citationsRead paper

SPEC CPU: The Next Generation

May 02, 2026

This work addresses the limitations of existing CPU benchmarks in accurately evaluating the performance of modern heterogeneous, multithreaded processors under diverse workloads. To this end, the authors present the SPEC CPU 2026 benchmark suite, developed through community collaboration and principled methodology, which introduces the Rolling-Round-Robin Rate approach to standardize the execution of heterogeneous multiprogrammed workloads. The suite incorporates newly designed multithreaded benchmarks exhibiting varied microarchitectural characteristics, selected and hardened through an open-source application curation process. Emphasizing workload diversity, portability, and long-term viability, SPEC CPU 2026 establishes a robust, representative, and authoritative standard for performance evaluation, thereby supporting next-generation computer architecture research.

0 citationsRead paper

I hope we don't do to trust what advertising has done to love

Apr 30, 2026

This work addresses the prevalent misuse and conceptual vagueness surrounding “trust” in contemporary artificial intelligence, particularly within agent-based systems—a phenomenon akin to the commodification of “love” in advertising. The paper introduces, for the first time, a structured “pillars of trust” framework that operationalizes abstract notions of trust into measurable, actionable dimensions. Through human-AI interaction design, these dimensions are concretized as a “trust vector.” This framework not only facilitates interdisciplinary dialogue across computational domains and civil society but also establishes a theoretical foundation and public deliberation mechanism for the design and evaluation of trustworthy AI systems, thereby advancing the standardization and transparency of trust in technical practice.

0 citationsRead paper

Crystal structure prediction using graph neural combinatorial optimization

Apr 26, 2026

This work addresses the discrete combinatorial optimization challenge in crystal structure prediction arising from the absence of symmetry constraints by introducing, for the first time, graph neural combinatorial optimization to this domain. The proposed method constructs an extended graph to simultaneously model both short-range and long-range atomic interactions and integrates the Gumbel-Sinkhorn mechanism to efficiently generate crystal structures that satisfy a target stoichiometry within an unsupervised framework. Evaluated across diverse chemical compositions, the approach significantly outperforms classical heuristic algorithms and achieves performance comparable to commercial optimization solvers, thereby establishing a new paradigm for GPU-accelerated, large-scale crystal structure prediction.

0 citationsRead paper