Large language models (LLMs) deliver impressive performance but require large amounts of energy. In this work, we present a MatMul-free LLM architecture adapted for Intel's neuromorphic processor, Loihi 2. Our approach leverages Loihi 2's support for low-precision, event-driven computation and stateful processing. Our hardware-aware quantized model on GPU demonstrates that a 370M parameter MatMul-free model can be quantized with no accuracy loss. Based on preliminary results, we report up to 3x higher throughput with 2x less energy, compared to transformer-based LLMs on an edge GPU, with significantly better scaling. Further hardware optimizations will increase throughput and decrease energy consumption. These results show the potential of neuromorphic hardware for efficient inference and pave the way for efficient reasoning models capable of generating complex, long-form text rapidly and cost-effectively.

Linear recurrent neural networks enable powerful long-range sequence modeling with constant memory usage and time-per-token during inference. These architectures hold promise for streaming applications at the edge, but deployment in resource-constrained environments requires hardware-aware optimizations to minimize latency and energy consumption. Unstructured sparsity offers a compelling solution, enabling substantial reductions in compute and memory requirements--when accelerated by compatible hardware platforms. In this paper, we conduct a scaling study to investigate the Pareto front of performance and efficiency across inference compute budgets. We find that highly sparse linear RNNs consistently achieve better efficiency-performance trade-offs than dense baselines, with 2x less compute and 36% less memory at iso-accuracy. Our models achieve state-of-the-art results on a real-time streaming task for audio denoising. By quantizing our sparse models to fixed-point arithmetic and deploying them on the Intel Loihi 2 neuromorphic chip for real-time processing, we translate model compression into tangible gains of 42x lower latency and 149x lower energy consumption compared to a dense model on an edge GPU. Our findings showcase the transformative potential of unstructured sparsity, paving the way for highly efficient recurrent neural networks in real-world, resource-constrained environments.

Large language models have transformed AI, yet reliably controlling their outputs remains a challenge. This paper explores activation engineering, where outputs of pre-trained LLMs are controlled by manipulating their activations at inference time. Unlike traditional methods using a single steering vector, we introduce conceptors - mathematical constructs that represent sets of activation vectors as ellipsoidal regions. Conceptors act as soft projection matrices and offer more precise control over complex activation patterns. Our experiments demonstrate that conceptors outperform traditional methods across multiple steering tasks. We further use Boolean operations on conceptors for combined steering goals that empirically outperform additively combining steering vectors on a set of tasks. These results highlight conceptors as a promising tool for more effective steering of LLMs. Our code is available on github.com/jorispos/conceptorsteering.

The value of brain-inspired neuromorphic computers critically depends on our ability to program them for relevant tasks. Currently, neuromorphic hardware often relies on machine learning methods adapted from deep learning. However, neuromorphic computers have potential far beyond deep learning if we can only harness their energy efficiency and full computational power. Neuromorphic programming will necessarily be different from conventional programming, requiring a paradigm shift in how we think about programming. This paper presents a conceptual analysis of programming within the context of neuromorphic computing, challenging conventional paradigms and proposing a framework that aligns more closely with the physical intricacies of these systems. Our analysis revolves around five characteristics that are fundamental to neuromorphic programming and provides a basis for comparison to contemporary programming methods and languages. By studying past approaches, we contribute a framework that advocates for underutilized techniques and calls for richer abstractions to effectively instrument the new hardware class.

Modern recurrent layers are emerging as a promising path toward edge deployment of foundation models, especially in the context of large language models (LLMs). Compressing the whole input sequence in a finite-dimensional representation enables recurrent layers to model long-range dependencies while maintaining a constant inference cost for each token and a fixed memory requirement. However, the practical deployment of LLMs in resource-limited environments often requires further model compression, such as quantization and pruning. While these techniques are well-established for attention-based models, their effects on recurrent layers remain underexplored. In this preliminary work, we focus on post-training quantization for recurrent LLMs and show that Mamba models exhibit the same pattern of outlier channels observed in attention-based LLMs. We show that the reason for the difficulty of quantizing SSMs is caused by activation outliers, similar to those observed in transformer-based LLMs. We report baseline results for post-training quantization of Mamba that do not take into account the activation outliers and suggest first steps for outlier-aware quantization.

In the quest for next-generation sequence modeling architectures, State Space Models (SSMs) have emerged as a potent alternative to transformers, particularly for their computational efficiency and suitability for dynamical systems. This paper investigates the effect of quantization on the S5 model to understand its impact on model performance and to facilitate its deployment to edge and resource-constrained platforms. Using quantization-aware training (QAT) and post-training quantization (PTQ), we systematically evaluate the quantization sensitivity of SSMs across different tasks like dynamical systems modeling, Sequential MNIST (sMNIST) and most of the Long Range Arena (LRA). We present fully quantized S5 models whose test accuracy drops less than 1% on sMNIST and most of the LRA. We find that performance on most tasks degrades significantly for recurrent weights below 8-bit precision, but that other components can be compressed further without significant loss of performance. Our results further show that PTQ only performs well on language-based LRA tasks whereas all others require QAT. Our investigation provides necessary insights for the continued development of efficient and hardware-optimized SSMs.

Machine learning has made tremendous progress in recent years and received large amounts of public attention. Though we are still far from designing a full artificially intelligent agent, machine learning has brought us many applications in which computers solve human learning tasks remarkably well. Much of this progress comes from a recent trend within machine learning, called deep learning. Deep learning models are responsible for many state-of-the-art applications of machine learning. Despite their success, deep learning models are hard to train, very difficult to understand, and often times so complex that training is only possible on very large GPU clusters. Lots of work has been done on enabling neural networks to learn efficiently. However, the design and architecture of such neural networks is often done manually through trial and error and expert knowledge. This thesis inspects different approaches, existing and novel, to automate the design of deep feedforward neural networks in an attempt to create less complex models with good performance that take away the burden of deciding on an architecture and make it more efficient to design and train such deep networks.