In this paper, we propose Patience-based Early Exit, a straightforward yet effective inference method that can be used as a plug-and-play technique to simultaneously improve the efficiency and robustness of a pretrained language model (PLM). To achieve this, our approach couples an internal-classifier with each layer of a PLM and dynamically stops inference when the intermediate predictions of the internal classifiers do not change for a pre-defined number of steps. Our approach improves inference efficiency as it allows the model to make a prediction with fewer layers. Meanwhile, experimental results with an ALBERT model show that our method can improve the accuracy and robustness of the model by preventing it from overthinking and exploiting multiple classifiers for prediction, yielding a better accuracy-speed trade-off compared to existing early exit methods.

Currently, pre-trained models can be considered the default choice for a wide range of NLP tasks. Despite their SoTA results, there is practical evidence that these models may require a different number of computing layers for different input sequences, since evaluating all layers leads to overconfidence in wrong predictions (namely overthinking). This problem can potentially be solved by implementing adaptive computation time approaches, which were first designed to improve inference speed. Recently proposed PonderNet may be a promising solution for performing an early exit by treating the exit layer's index as a latent variable. However, the originally proposed exit criterion, relying on sampling from trained posterior distribution on the probability of exiting from the $i$-th layer, introduces major variance in exit layer indices, significantly reducing the resulting model's performance. In this paper, we propose improving PonderNet with a novel deterministic Q-exit criterion and a revisited model architecture. We adapted the proposed mechanism to ALBERT and RoBERTa and compared it with recent methods for performing an early exit. We observed that the proposed changes can be considered significant improvements on the original PonderNet architecture and outperform PABEE on a wide range of GLUE tasks. In addition, we also performed an in-depth ablation study of the proposed architecture to further understand Lambda layers and their performance.

Early-Exit Large Language Models (EE-LLMs) enable high throughput inference by allowing tokens to exit early at intermediate layers. However, their throughput is limited by the computational and memory savings. Existing EE-LLM frameworks rely on a single model and therefore, their token generation latencies are bottlenecked by tokens that do not exit early and traverse additional layers. Moreover, early exits are only known at runtime and depend on the request. Therefore, these frameworks load the weights of all model layers even though large portions remain unused when tokens exit early. The lack of memory savings limit us from scaling the batch sizes. We propose $\textit{HELIOS}$, a framework that improves both token generation latency and batch sizes to enable high-throughput in EE-LLMs. HELIOS exploits two insights. $\textit{First}$, early exits are often complimentary across models, tokens that do not exit early on one model often take an early-exit on another. HELIOS employs multiple models and dynamically switches between them to collectively maximize the number of tokens that exit early, and minimize token generation latencies. $\textit{Second}$, even when a predicted token does not exit early due to poor confidence, it often remains unchanged even after additional layer traversal. HELIOS greedily allows such tokens to exit early and only loads the weights of the most likely to be used layers, yielding memory savings which is then re-purposed to increase batch sizes. HELIOS employs real-time profiling to accurately identify the early-exit distributions, and adaptively switches between models by tracking tokens in real-time to minimize the performance degradation caused by greedy model loading and exiting. Our evaluations show that HELIOS achieves $1.48\times$ higher throughput and $15.14\times$ larger batch size compared to existing EE-LLM frameworks.

Reasoning models have demonstrated exceptional performance in tasks such as mathematics and logical reasoning, primarily due to their ability to engage in step-by-step thinking during the reasoning process. However, this often leads to overthinking, resulting in unnecessary computational overhead. To address this issue, Mode Selection aims to automatically decide between Long-CoT (Chain-of-Thought) or Short-CoT by utilizing either a Thinking or NoThinking mode. Simultaneously, Early Exit determines the optimal stopping point during the iterative reasoning process. Both methods seek to reduce the computational burden. In this paper, we first identify Mode Selection as a more challenging variant of the Early Exit problem, as they share similar objectives but differ in decision timing. While Early Exit focuses on determining the best stopping point for concise reasoning at inference time, Mode Selection must make this decision at the beginning of the reasoning process, relying on pre-defined fake thoughts without engaging in an explicit reasoning process, referred to as zero-step thinking. Through empirical studies on nine baselines, we observe that prompt-based approaches often fail due to their limited classification capabilities when provided with minimal hand-crafted information. In contrast, approaches that leverage internal information generally perform better across most scenarios but still exhibit issues with stability. Our findings indicate that existing methods relying solely on the information provided by models are insufficient for effectively addressing Mode Selection in scenarios with limited information, highlighting the ongoing challenges of this task. Our code is available at https://github.com/Trae1ounG/Zero_Step_Thinking.

The Tsetlin Machine (TM) offers high-speed inference on resource-constrained devices such as CPUs. Its logic-driven operations naturally lend themselves to parallel execution on modern CPU architectures. Motivated by this, we propose an efficient software implementation of the TM by leveraging instruction-level bitwise operations for compact model representation and accelerated processing. To further improve inference speed, we introduce an early exit mechanism, which exploits the TM's AND-based clause evaluation to avoid unnecessary computations. Building upon this, we propose a literal Reorder strategy designed to maximize the likelihood of early exits. This strategy is applied during a post-training, pre-inference stage through statistical analysis of all literals and the corresponding actions of their associated Tsetlin Automata (TA), introducing negligible runtime overhead. Experimental results using the gem5 simulator with an ARM processor show that our optimized implementation reduces inference time by up to 96.71% compared to the conventional integer-based TM implementations while maintaining comparable code density.

The Efficient Adaptive Transformer (EAT) framework unifies three adaptive efficiency techniques - progressive token pruning, sparse attention, and dynamic early exiting - into a single, reproducible architecture for input-adaptive inference. EAT provides an open-source benchmarking pipeline that automates data processing, timing, and ablation across GLUE tasks (SST-2, QQP, MNLI). Although this empirical study finds that combining these mechanisms can increase latency in shallow six-layer models, it demonstrates that EAT achieves slightly higher accuracy than the optimized DistilBERT baseline on SST-2, illustrating the potential of dynamic computation for latency-sensitive NLP. The main contribution is the open, end-to-end reproducible framework - complete with scripts, CSV logging, and analysis utilities - intended to serve as a community tool for further research on adaptive transformers.