However, one might argue that this analysis might not allow for sufficient differentiation between tasks. To address this concern, we expanded our evaluation to the entire MMLU benchmark. This enabled a comparable assessment of task similarity, akin to our earlier experiments.

What does it truly mean for a language model to "reason"? Most current evaluations and benchmarks reward models' correct standalone answers--but correctness alone reveals little about the process that produced them. In this work, we explore a different perspective: reasoning is not a static chain of steps, but a dynamic trajectory where ideas interact, clash, and evolve into deeper insights. To capture this dynamic, we draw on a well-established philosophical tradition: dialectics, where reasoning unfolds through thesis, antithesis, and synthesis. Building on this, we present SIEV, a structured framework that evaluates reasoning of LLMs through dialectics. Unlike conventional evaluations, SIEV assesses not only the conclusion a model reaches, but how it gets there: its ability to resolve tension, integrate distinct ideas, and synthesize higher-order reasoning. This lens uncovers significant reasoning gaps in state-of-the-art models even under saturated benchmarks like GSM and MMLU. For instance, GPT -5-chat, a recent model, loses over 40 points (out of 100) when evaluated with SIEV on GSM. Our findings highlight that adopting a process-oriented, philosophically grounded approach enables a deeper, more rigorous, and more discriminative assessment of LLM reasoning. Reasoning and LLMs: Reasoning is central to how people solve problems and make decisions, and it is increasingly vital for LLMs in real-world use. Traditionally, LLM performance has been assessed using benchmarks that span diverse domains (e.g., GPQA Rein et al. (2023), MMLU-Pro Wang et al. (2024), AIME HuggingFaceH4 (2024), etc.). While these benchmarks offer various metrics to cover comparing models in wide range of topics, the core evaluation paradigm remains largely unchanged: did the model get the right answer? We argue that this narrow focus only on the direct standalone responses is increasingly inadequate--especially when evaluating reasoning. It overlooks the depth, robustness, and coherence of the reasoning process itself. To address this, a shift toward evaluating how models reason--not just what they conclude--is needed.

For question-answering (QA) tasks, in-context learning (ICL) enables language models to generate responses without modifying their parameters by leveraging examples provided in the input. However, the effectiveness of ICL heavily depends on the availability of high-quality examples, which are often scarce due to data privacy constraints, annotation costs, and distribution disparities. A natural solution is to utilize examples stored on client devices, but existing approaches either require transmitting model parameters - incurring significant communication overhead - or fail to fully exploit local datasets, limiting their effectiveness. To address these challenges, we propose Federated In-Context Learning (Fed-ICL), a general framework that enhances ICL through an iterative, collaborative process. Fed-ICL progressively refines responses by leveraging multi-round interactions between clients and a central server, improving answer quality without the need to transmit model parameters. We establish theoretical guarantees for the convergence of Fed-ICL and conduct extensive experiments on standard QA benchmarks, demonstrating that our proposed approach achieves strong performance while maintaining low communication costs.

Multi-agent systems address issues of accessibility and scalability of artificial intelligence (AI) foundation models, which are often represented by large language models. We develop a framework - the "Society of HiveMind" (SOHM) - that orchestrates the interaction between multiple AI foundation models, imitating the observed behavior of animal swarms in nature by following modern evolutionary theories. On the one hand, we find that the SOHM provides a negligible benefit on tasks that mainly require real-world knowledge. On the other hand, we remark a significant improvement on tasks that require intensive logical reasoning, indicating that multi-agent systems are capable of increasing the reasoning capabilities of the collective compared to the individual agents. Our findings demonstrate the potential of combining a multitude of diverse AI foundation models to form an artificial swarm intelligence capable of self-improvement through interactions with a given environment.

Recently, Large Language Models (LLMs) with Mixture of Experts (MoE) layers have gained significant attention. Currently, state-of-the-art LLMs utilize this architecture. There is a substantial amount of research on how to train such models and how to select hyperparameters for this architecture. However, there is a lack of studies focusing on post-evaluation analysis of MoE layer properties. In this paper, we take a first step toward closing this gap by evaluating expert contributions on the quiz-based MMLU benchmark. We show that most experts were never activated during inference on this benchmark. Additionally, the output distribution of gating networks is much closer to uniform than sparse. Finally, we demonstrate that the average performance of some experts within the same layer varies significantly.

In this paper, we propose an architecture to harness the collective knowledge of multiple trained LLMs to create a new state-of-the-art. At the core of this framework is a LLM-based orchestrator that is adept at picking the right underlying LLM experts for optimal task execution. Inspired by self-play in reinforcement learning, we created a loop of query generation, orchestration, and evaluation to generate training data for the orchestrator. Our evaluation focused on the MMLU benchmark, employing models with 7B, 13B, and 34B parameters available on Hugging Face. The results demonstrate new state-of-the-art open-source models: Our Leeroo orchestrator achieves performance on par with the Mixtral model while incurring only two-thirds of its cost. Moreover, increasing the allowed cost surpasses Mixtral's accuracy by over 5% at the same cost level, reaching an accuracy of 75.9%. Further enhancements were observed when integrating GPT4 into the underlying model pool. The Leeroo orchestrator nearly matches GPT4's performance at half the cost and even exceeds GPT4's results with a 25% cost reduction. These findings illustrate the potential of our architecture in creating state-of-the-art and cost-effective LLMs by optimizing the synergy between multiple LLMs to achieve superior performance outcomes.