This paper introduces a structured memory which can be easily integrated into a neural network. The memory is very large by design and significantly increases the capacity of the architecture, by up to a billion parameters with a negligible computational overhead.

However, the model size and corresponding computational complexity are constantly scaled up in pursuit of higher performance.

This paper introduces a structured memory which can be easily integrated into a neural network. The memory is very large by design and significantly increases the capacity of the architecture, by up to a billion parameters with a negligible computational overhead. Its design and access pattern is based on product keys, which enable fast and exact nearest neighbor search. The ability to increase the number of parameters while keeping the same computational budget lets the overall system strike a better trade-off between prediction accuracy and computation efficiency both at training and test time. This memory layer allows us to tackle very large scale language modeling tasks. In our experiments we consider a dataset with up to 30 billion words, and we plug our memory layer in a state-of-the-art transformer-based architecture. In particular, we found that a memory augmented model with only 12 layers outperforms a baseline transformer model with 24 layers, while being twice faster at inference time. We release our code for reproducibility purposes.

The transition from human-centric to agent-centric software development practices is disrupting existing knowledge sharing environments for software developers. Traditional peer-to-peer repositories and developer communities for shared technical knowledge and best practice have witnessed dramatic drops in participation in a short period of time. At the same time, agentic functional equivalents are yet to emerge leaving AI agents, which already generate a significant proportion of all new software code produced, without access to repositories of valuable shared learning. In this paper, we introduce Spark, a novel shared agentic memory architecture which is designed to emulate the collective intelligence and know-how of human developer communities. Spark enables AI coding agents to both contribute to and draw from a persistent and continuously evolving experiential memory. Agents operating in the same general problem space use the Spark shared memory as a repository of new knowledge to achieve collective continual learning. We evaluate Spark as a coach for AI coding agents performing software development tasks. We demonstrate that recommendations made by Spark improve the quality of code generated by generic code generation models at varying sizes and capability tiers. Boosted by Spark, a small open-weights model with 30 billion parameters was able to match the code quality afforded by a much larger state-of-the-art model. Separately, we measure the intrinsic quality of recommendations generated by Spark against a wide range of criteria inspired by software development best practice, and achieve helpfulness levels of up to 98.2% in the top two (out of five) qualitative helpfulness bands.

Modern language models are powerful, but typically static after deployment. A major obstacle to building models that continually learn over time is catastrophic forgetting, where updating on new data erases previously acquired capabilities. Motivated by the intuition that mitigating forgetting is challenging because trainable parameters are shared across all tasks, we investigate whether sparse parameter updates can enable learning without catastrophic forgetting. We introduce sparse memory finetuning, leveraging memory layer models (Berges et al., 2024), which are sparsely updated by design. By updating only the memory slots that are highly activated by a new piece of knowledge relative to usage on pretraining data, we reduce interference between new knowledge and the model's existing capabilities. We evaluate learning and forgetting compared to full finetuning and parameter-efficient finetuning with LoRA on two question answering tasks. We find that sparse memory finetuning learns new knowledge while exhibiting substantially less forgetting: while NaturalQuestions F1 drops by 89% after full finetuning on new facts and 71% with LoRA, sparse memory finetuning yields only an 11% drop with the same level of new knowledge acquisition. Our results suggest sparsity in memory layers offers a promising path toward continual learning in large language models.

However, the model size and corresponding computational complexity are constantly scaled up in pursuit of higher performance.

Neural Information Processing Systems http://nips.cc/

While Mixture of Experts (MoE) models achieve remarkable efficiency by activating only subsets of parameters, they suffer from high memory access costs during inference. Memory-layer architectures offer an appealing alternative with very few memory access, but previous attempts like UltraMem have only matched the performance of 2-expert MoE models, falling significantly short of state-of-the-art 8-expert configurations. We present UltraMemV2, a redesigned memory-layer architecture that closes this performance gap. Our approach introduces five key improvements: integrating memory layers into every transformer block, simplifying value expansion with single linear projections, adopting FFN-based value processing from PEER, implementing principled parameter initialization, and rebalancing memory-to-FFN computation ratios. Through extensive evaluation, we demonstrate that UltraMemV2 achieves performance parity with 8-expert MoE models under same computation and parameters but significantly low memory access. Notably, UltraMemV2 shows superior performance on memory-intensive tasks, with improvements of +1.6 points on long-context memorization, +6.2 points on multi-round memorization, and +7.9 points on in-context learning. We validate our approach at scale with models up to 2.5B activated parameters from 120B total parameters, and establish that activation density has greater impact on performance than total sparse parameter count. Our work brings memory-layer architectures to performance parity with state-of-the-art MoE models, presenting a compelling alternative for efficient sparse computation.

Shoaib Jameel University of Southampton Southampton, United Kingdom M.S.Jameel@southampton.ac.uk Abstract --Personality detection automatically identifies an individual's personality from various data sources, such as social media texts. However, as the parameter scale of language models continues to grow, the computational cost becomes increasingly difficult to manage. Fine-tuning also grows more complex, making it harder to justify the effort and reliably predict outcomes. We introduce a novel parameter-efficient fine-tuning framework, PersLLM, to address these challenges. By storing the features in the memory layer, we eliminate the need for repeated complex computations by the LLM. Meanwhile, the lightweight output network serves as a proxy for evaluating the overall effectiveness of the framework, improving the predictability of results. Experimental results on key benchmark datasets like Kaggle and Pandora show that PersLLM significantly reduces computational cost while maintaining competitive performance and strong adaptability. I NTRODUCTION Personality refers to the stable traits in an individual's emotions, thoughts, and behaviours that shape how they perceive, interpret, and interact with the world [1], [2]. It is a complex and multifaceted construct that combines various characteristics to form a person's unique identity.

UPDATE: Authors answered my questions, I would like to keep my score unchanged and suggest to focus on clarity of the final version. Perhaps, this is the case when I would really be interested in looking at the source code. Originality: the paper borrows the general idea of product keys from the database community, however the application to fast retrieval in neural memory systems seems quite novel to me. Quality: The core ideas of the paper are sound, however more I would appreciate more rigor in both conceptual and experimental comparison with other approaches incorporating memory to Transformer (see e.g. Another suggestion would be to discuss more the issue of potential non-uniformity of the query distribution, which indeed seems to be quite relevant.