University of the Immaculate Conception

In the United States, pancreatic cancer ranks in third place amongst all cancers, with a very low survival rate of 20

Research in artificial intelligence is shifting from model innovations and benchmark scores towards problem definition and rigorous real-world evaluation. As the field enters the "second half," the central challenge becomes real utility in long-horizon, dynamic, and user-dependent settings such as agentic coding, deep research, and computer use, where LLM-based agents face context explosion beyond fixed context windows and must continuously accumulate, manage, and selectively reuse information across extended interactions. Memory, with hundreds of papers released in 2025, therefore emerges as the critical solution to fill this utility gap. Beyond passive storage, memory is increasingly the substrate through which agents self-evolve: short-term memory gates which experiences are perceived and abstracted during execution, while long-term memory consolidates them into reusable knowledge and skills, forming the loop through which agents improve from their own experience. In this survey, we provide a unified view of foundation agent memory along three dimensions: memory substrate (internal parametric state and external retrieval-augmented stores), cognitive mechanism (sensory, working, episodic, semantic, and procedural), and memory subject (user-centric personalization and agent-centric experience). We then analyze how memory is operated under single- and multi-agent topologies and highlight learning policies over memory operations, showing how memory management itself is becoming a trainable capability spanning reinforcement-learned context curation, experience consolidation at decision time, and the emerging ecosystem of portable, shareable agent skills. Finally, we review evaluation benchmarks and metrics for memory utility, and outline open challenges and future directions.

Knowledge distillation from Large Language Models (LLMs) to smaller models has emerged as a critical technique for deploying efficient AI systems. However, current methods for distillation via synthetic data lack pedagogical awareness, treating knowledge transfer as a one-off data synthesis and training task rather than a systematic learning process. In this paper, we propose a novel pedagogically-inspired framework for LLM knowledge distillation that draws from fundamental educational principles. Our approach introduces a three-stage pipeline—**Knowledge Identifier**, **Organizer**, and **Adapter** (**IOA**)—that systematically identifies knowledge deficiencies in student models, organizes knowledge delivery through progressive curricula, and adapts representations to match the cognitive capacity of student models. We integrate Bloom's Mastery Learning Principles and Vygotsky's Zone of Proximal Development to create a dynamic distillation process where student models approach teacher model's performance on prerequisite knowledge before advancing, and new knowledge is introduced with controlled, gradual difficulty increments. Extensive experiments using LLaMA-3.1/3.2 and Qwen2.5 as student models demonstrate that IOA achieves significant improvements over baseline distillation methods, with student models retaining 94.7\% of teacher performance on DollyEval while using less than 1/10th of the parameters. Our framework particularly excels in complex reasoning tasks, showing 19.2\% improvement on MATH and 22.3\% on HumanEval compared with state-of-the-art baselines.

As large language models span dense, mixture-of-experts, and state-space architectures and are deployed on heterogeneous accelerators under increasingly diverse multimodal workloads, optimising inference energy has become as critical as optimizing latency and throughput. Existing approaches either treat latency as an energy proxy or rely on data-hungry black-box surrogates. Both fail under varying parallelism strategies: latency and energy optima diverge in over 20

Model merging in the pretrain-finetune paradigm has proven effective by combining multiple finetuned models into one with multi-task capabilities. Recent merging methods aim to boost merged models’ performance through strategies such as mitigating conflicts, adding trainable modules, and incorporating task-specific components. In most methods, the parameter merging procedure is based on Task Arithmetic, a widely used technique that creates task vectors from each finetuned model and linearly combines them with coefficients into consolidated model parameters. Except for studies specifically focusing on the merging coefficients, many other methods treat them as hand-tuned hyperparameters. However, the merging coefficients, which govern the entire merging process, including the subsequent module training, are empirically crucial for achieving optimal performance and tradeoff across tasks. Thus, this paper proposed an innovative model merging framework called MetaMerging, which constructs the merged model with a unified model and lightweight task-specific adapters. Specifically, the adapters are efficiently trained without labels via feature alignment with fine-tuned models, while the unified model is obtained by merging task vectors with coefficients adaptively optimized through meta-learning, which enhances the generalization and enables more effective adapter training. Extensive experiments on CV and NLP fields show strong performance of MetaMerging on various downstream tasks and demonstrate the effectiveness of meta-learning in our method compared to other parameter merging methods. Our code is available at https://anonymous.4open.science/r/MetaMerging-53A1