arXiv subjects

Shay Cohen

Publications and source records attributed to Shay Cohen.

5 recordsLinked to original sources

arXiv2026

Magenta: Closing the Loop Between Mathematical Reasoning and Lean Verification

Most of mathematical knowledge has been communicated through so-called informal use of mathematics and natural language. With large language models (LLMs) being highly adept in using natural language, they achieve strong performance, yet not perfect, in informal mathematical reasoning. Restraining LLMs to informal reasoning misses out on the opportunity to use the discrete verification abilities that machines offer through machine-checkable proofs. In this paper, we bridge the gap between informal and formal reasoning by integrating Lean signals into the informal reasoning process. We introduce Magenta, a training-free agentic pipeline that, given only a natural-language problem, produces an answer, expresses it as a Lean 4 statement, and constructs a machine-checked proof. A statement judge verifies whether the formalisation preserves the original problem, while an error-attribution judge routes failed attempts either to mathematical re-derivation or local Lean repair. Magenta achieves 100% accuracy across all evaluated olympiad benchmarks, including AIME 2025, AIME 2026, and HMMT February 2026. When paired with the open-weight K2-Horizon-7B reasoner, it solves all six IMO 2026 problems. Our analysis shows that statement adjudication is essential for preventing false certificates and that feedback-guided correction outperforms independent resampling on difficult problems.

cs.AI↗

arXiv2023

Neuron to Graph: Interpreting Language Model Neurons at Scale

Advances in Large Language Models (LLMs) have led to remarkable capabilities, yet their inner mechanisms remain largely unknown. To understand these models, we need to unravel the functions of individual neurons and their contribution to the network. This paper introduces a novel automated approach designed to scale interpretability techniques across a vast array of neurons within LLMs, to make them more interpretable and ultimately safe. Conventional methods require examination of examples with strong neuron activation and manual identification of patterns to decipher the concepts a neuron responds to. We propose Neuron to Graph (N2G), an innovative tool that automatically extracts a neuron's behaviour from the dataset it was trained on and translates it into an interpretable graph. N2G uses truncation and saliency methods to emphasise only the most pertinent tokens to a neuron while enriching dataset examples with diverse samples to better encompass the full spectrum of neuron behaviour. These graphs can be visualised to aid researchers' manual interpretation, and can generate token activations on text for automatic validation by comparison with the neuron's ground truth activations, which we use to show that the model is better at predicting neuron activation than two baseline methods. We also demonstrate how the generated graph representations can be flexibly used to facilitate further automation of interpretability research, by searching for neurons with particular properties, or programmatically comparing neurons to each other to identify similar neurons. Our method easily scales to build graph representations for all neurons in a 6-layer Transformer model using a single Tesla T4 GPU, allowing for wide usability. We release the code and instructions for use at https://github.com/alexjfoote/Neuron2Graph.

cs.LG↗

arXiv2023

Erasure of Unaligned Attributes from Neural Representations

We present the Assignment-Maximization Spectral Attribute removaL (AMSAL) algorithm, which erases information from neural representations when the information to be erased is implicit rather than directly being aligned to each input example. Our algorithm works by alternating between two steps. In one, it finds an assignment of the input representations to the information to be erased, and in the other, it creates projections of both the input representations and the information to be erased into a joint latent space. We test our algorithm on an extensive array of datasets, including a Twitter dataset with multiple guarded attributes, the BiasBios dataset and the BiasBench benchmark. The last benchmark includes four datasets with various types of protected attributes. Our results demonstrate that bias can often be removed in our setup. We also discuss the limitations of our approach when there is a strong entanglement between the main task and the information to be erased.

cs.CL↗

arXiv2020

Tensors over Semirings for Latent-Variable Weighted Logic Programs

Semiring parsing is an elegant framework for describing parsers by using semiring weighted logic programs. In this paper we present a generalization of this concept: latent-variable semiring parsing. With our framework, any semiring weighted logic program can be latentified by transforming weights from scalar values of a semiring to rank-n arrays, or tensors, of semiring values, allowing the modelling of latent variables within the semiring parsing framework. Semiring is too strong a notion when dealing with tensors, and we have to resort to a weaker structure: a partial semiring. We prove that this generalization preserves all the desired properties of the original semiring framework while strictly increasing its expressiveness.

cs.CL↗

arXiv2015

Coping with Space Neophobia in Drosophila melanogaster: The Asymmetric Dynamics of Crossing a Doorway to the Untrodden

Insects exhibit remarkable cognitive skills in the field and several cognitive abilities have been demonstrated in Drosophila in the laboratory. By devising an ethologically relevant experimental setup that also allows comparison of behavior across remote taxonomic groups we sought to reduce the gap between the field and the laboratory, and reveal as yet undiscovered ethological phenomena within a wider phylogenetic perspective. We tracked individual flies that eclosed in a small (45mm) arena containing a piece of fruit, connected to a larger (130mm) arena by a wide (5mm) doorway. Using this setup we show that the widely open doorway initially functions as a barrier: the likelihood of entering the large arena increases gradually, requiring repeated approaches to the doorway, and even after entering the flies immediately return. Gradually the flies acquire the option to avoid returning, spending more relative time and performing relatively longer excursions in the large arena. The entire process may take up three successive days. This behavior constitutes coping with space neophobia, the avoidance of untrodden space. It appears to be the same as the neophobic doorway-crossing reported in mouse models of anxiety. In both mice and flies the moment-to-moment developmental dynamics of transition between trodden and untrodden terrain appear to be the same, and in mice it is taken to imply memory and, therefore, cognition. Recent claims have been made for a deep homology between the arthropod central complex and the vertebrate basal ganglia, two structures involved in navigation. The shared dynamics of space occupancy in flies and mice might indicate the existence of cognitive exploration also in the flies or else a convergent structure exhibiting the same developmental dynamics.

q-bio.NC↗