Answering complex real-world questions requires step-by-step retrieval and integration of relevant information to generate well-grounded responses. However, existing knowledge distillation methods overlook the need for different reasoning abilities at different steps, hindering transfer in multi-step retrieval-augmented frameworks. To address this, we propose Stepwise Knowledge Distillation for Enhancing Reasoning Ability in Multi-Step Retrieval-Augmented Language Models (StepER). StepER employs step-wise supervision to align with evolving information and reasoning demands across stages. Additionally, it incorporates difficulty-aware training to progressively optimize learning by prioritizing suitable steps. Our method is adaptable to various multi-step retrieval-augmented language models, including those that use retrieval queries for reasoning paths or decomposed questions. Extensive experiments show that StepER outperforms prior methods on multi-hop QA benchmarks, with an 8B model achieving performance comparable to a 70B teacher model.

Recent advances in large language models (LLMs) have introduced the novel paradigm of using LLMs as judges, where an LLM evaluates and scores the outputs of another LLM, which often correlates highly with human preferences. However, the use of LLM-as-a-judge has been primarily studied in English. In this paper, we evaluate this framework in Russian by introducing the Russian Error tyPes Annotation dataset (REPA), a dataset of 1k user queries and 2k LLM-generated responses. Human annotators labeled each response pair expressing their preferences across ten specific error types, as well as selecting an overall preference. We rank six generative LLMs across the error types using three rating systems based on human preferences. We also evaluate responses using eight LLM judges in zero-shot and few-shot settings. We describe the results of analyzing the judges and position and length biases. Our findings reveal a notable gap between LLM judge performance in Russian and English. However, rankings based on human and LLM preferences show partial alignment, suggesting that while current LLM judges struggle with fine-grained evaluation in Russian, there is potential for improvement.

United States defence secretary Lloyd Austin has waded in on reports that North Korea was preparing to enter the Ukraine war with troops. "If they are co-belligerents, if their intention is to participate in this war on Russia's behalf, that is a very, very serious issue," Austin said. Austin was returning from his fourth visit to Kyiv, where he announced a 400m package of US weapons for Ukraine. John Kirby, White House national security spokesman, said Washington believes that at least 3,000 North Korean soldiers arrived this month by sea to Vladivostok, Russia's largest Pacific port. "These soldiers then travelled onward to multiple Russian military training sites in eastern Russia, where they are currently undergoing training," Kirby said on Wednesday.

We have developed a reinforcement learning agent that often finds a minimal sequence of unknotting crossing changes for a knot diagram with up to 200 crossings, hence giving an upper bound on the unknotting number. We have used this to determine the unknotting number of 57k knots. We took diagrams of connected sums of such knots with oppositely signed signatures, where the summands were overlaid. The agent has found examples where several of the crossing changes in an unknotting collection of crossings result in hyperbolic knots. Based on this, we have shown that, given knots $K$ and $K'$ that satisfy some mild assumptions, there is a diagram of their connected sum and $u(K) + u(K')$ unknotting crossings such that changing any one of them results in a prime knot. As a by-product, we have obtained a dataset of 2.6 million distinct hard unknot diagrams; most of them under 35 crossings. Assuming the additivity of the unknotting number, we have determined the unknotting number of 43 at most 12-crossing knots for which the unknotting number is unknown.

Prompting-based large language models (LLMs) are surprisingly powerful at generating natural language reasoning steps or Chains-of-Thoughts (CoT) for multi-step question answering (QA). They struggle, however, when the necessary knowledge is either unavailable to the LLM or not up-to-date within its parameters. While using the question to retrieve relevant text from an external knowledge source helps LLMs, we observe that this one-step retrieve-and-read approach is insufficient for multi-step QA. Here, \textit{what to retrieve} depends on \textit{what has already been derived}, which in turn may depend on \textit{what was previously retrieved}. To address this, we propose IRCoT, a new approach for multi-step QA that interleaves retrieval with steps (sentences) in a CoT, guiding the retrieval with CoT and in turn using retrieved results to improve CoT. Using IRCoT with GPT3 substantially improves retrieval (up to 21 points) as well as downstream QA (up to 15 points) on four datasets: HotpotQA, 2WikiMultihopQA, MuSiQue, and IIRC. We observe similar substantial gains in out-of-distribution (OOD) settings as well as with much smaller models such as Flan-T5-large without additional training. IRCoT reduces model hallucination, resulting in factually more accurate CoT reasoning. Code, data, and prompts are available at \url{https://github.com/stonybrooknlp/ircot}

In this article, we propose the approach to structural optimization of neural networks, based on the braid theory. The paper describes the basics of braid theory as applied to the description of graph structures of neural networks. It is shown how networks of various topologies can be built using braid structures between layers of neural networks. The operation of a neural network based on the braid theory is compared with a homogeneous deep neural network and a network with random intersections between layers that do not correspond to the ordering of the braids. Results are obtained showing the advantage of braid-based networks over comparable architectures in classification problems.

The architecture of neural networks is selected by studying their accuracy and ability of generalization. This approach is not optimal and requires a lot of time and computational resources. So, it can be useful to replace it by the automatic optimization of neural network architectures. Automatic approaches imply the use of algorithmic (procedural) methods for the generation of neural networks, that is, the application of rules and procedures that create certain [1] sequences. This approach will be useful for generating deep neural network architectures, since the optimal setting of their structure is nontrivial and directly depends on the problem being solved.

Parallels between the signal processing tasks and biological neurons lead to an understanding of the principles of self-organized optimization of input signal recognition. In the present paper, we discuss such similarities among biological and technical systems. We propose the addition to the well-known STDP synaptic plasticity rule to directs the weight modification towards the state associated with the maximal difference between the background noise and correlated signals. The principle of physically constrained weight growth is used as a basis for such control of the modification of the weights. It is proposed, that biological synaptic straight modification is restricted by the existence and production of bio-chemical 'substances' needed for plasticity development. In this paper, the information about the noise-to-signal ratio is used to control such a substances' production and storage and to drive the neuron's synaptic pressures towards the state with the best signal-to-noise ratio. Several experiments with different input signal regimes are considered to understand the functioning of the proposed approach.

Biological neurons have adaptive nature and perform complex computations involving the filtering of redundant information. Such processing is often associated with Bayesian inference. Yet most common models of neural cells, including biologically plausible, such as Hodgkin-Huxley or Izhikevich do not possess predictive dynamics on the level of a single cell. The modern rules of synaptic plasticity or interconnections weights adaptation also do not provide grounding for the ability of neurons to adapt to the ever-changing input signal intensity. While natural neuron synaptic growth is precisely controlled and restricted by protein supply and recycling, weight correction rules such as widely used STDP are efficiently unlimited in change rate and scale. In the present article, we will introduce new mechanics of interconnection between neuron firing rate homeostasis and weight change by means of STDP growth bounded by abstract protein reserve, controlled by the intracellular optimization algorithm. We will show, how these cellular dynamics help neurons to filter out the intense signals to help neurons keep a stable firing rate. We will also examine that such filtering does not affect the ability of neurons to recognize the correlated inputs in unsupervised mode. Such an approach might be used in the machine learning domain to improve the robustness of AI systems. Modern neural networks and deep learning systems still lack the generalization and self-learning abilities of natural brains. Also, deep neural nets (DNN) need a lot of labeled data. Being tuned for one particular task and dataset DNNs may not perform so well in real practical application. These are major obstructions in the widespread implementation of deep learning systems for practical use [1]. Yet, training of SNN still needs to be improved to be widely used.