The transcriptional regulator Rbpj is involved in T-helper (T
) subset polarization, but its function in T
cells remains unclear. Here we show that T
-specific Rbpj deletion leads to splenomegaly and ...lymphadenopathy despite increased numbers of T
cells with a polyclonal TCR repertoire. A specific defect of Rbpj-deficient T
cells in controlling T
2 polarization and B cell responses is observed, leading to the spontaneous formation of germinal centers and a T
2-associated immunoglobulin class switch. The observed phenotype is environment-dependent and can be induced by infection with parasitic nematodes. Rbpj-deficient T
cells adopt open chromatin landscapes and gene expression profiles reminiscent of tissue-derived T
2-polarized T
cells, with a prevailing signature of the transcription factor Gata-3. Taken together, our study suggests that T
cells require Rbpj to specifically restrain T
2 responses, including their own excessive T
2-like differentiation potential.
We show that a purely dielectric structure made of alternating layers of deep subwavelength thicknesses exhibits novel transmission effects which completely contradict conventional effective medium ...theories exactly in the regime in which those theories are commonly used. We study waves incident at the vicinity of the effective medium's critical angle for total internal reflection and show that the transmission through the multilayer structure depends strongly on nanoscale variations even at layer thicknesses smaller than λ/50. In such deep subwavelength structures, we demonstrate dramatic changes in the transmission for variations in properties such as periodicity, order of the layers, and their parity. In addition to its conceptual importance, such sensitivity has important potential applications in sensing and switching.
We find that waves propagating in a 1D medium that is homogeneous in its linear properties but spatially disordered in its nonlinear coefficients undergo diffusive transport, instead of being ...Anderson localized as always occurs for linear disordered media. Specifically, electromagnetic waves in a multilayer structure with random nonlinear coefficients exhibit diffusion with features fundamentally different from the traditional diffusion in linear noninteracting systems. This unique transport, which stems from the nonlinear interaction between the waves and the disordered medium, displays anomalous statistical behavior where the fields in multiple different realizations converge to the same intensity value as they penetrate deeper into the medium.
Large language models (LLMs) are prone to hallucinations, which sparked a widespread effort to detect and prevent them. Recent work attempts to mitigate hallucinations by intervening in the model's ...generation, typically computing representative vectors of hallucinations vs. grounded generations, for steering the model's hidden states away from a hallucinatory state. However, common studies employ different setups and do not properly separate different possible causes of hallucinations, making interventions misguided. In this work, we introduce a method for categorizing examples based on the model's prior knowledge, named WACK. We construct WACK benchmarks that support interventions in two settings: open-book and closed-book question answering. Using the benchmarks, we perform an extensive investigation of the effect of different choices for intervention, such as the intervened components, and how often and how strongly to intervene. We find that intervention success varies depending on the component, with the attention blocks performing well and the residual stream proving detrimental to language modeling capabilities. We also show that interventions can benefit from representative vectors collected before, rather than after, a hallucination occurs. Finally, we introduce a new dynamic intervention, which intervenes only if needed, and thus is more robust than standard static interventions. The code is available at https://github.com/technion-cs-nlp/hallucination-mitigation .
Automatically determining whether a text and a corresponding image are semantically aligned is a significant challenge for vision-language models, with applications in generative text-to-image and ...image-to-text tasks. In this work, we study methods for automatic text-image alignment evaluation. We first introduce SeeTRUE: a comprehensive evaluation set, spanning multiple datasets from both text-to-image and image-to-text generation tasks, with human judgements for whether a given text-image pair is semantically aligned. We then describe two automatic methods to determine alignment: the first involving a pipeline based on question generation and visual question answering models, and the second employing an end-to-end classification approach by finetuning multimodal pretrained models. Both methods surpass prior approaches in various text-image alignment tasks, with significant improvements in challenging cases that involve complex composition or unnatural images. Finally, we demonstrate how our approaches can localize specific misalignments between an image and a given text, and how they can be used to automatically re-rank candidates in text-to-image generation.
Prompting language models to provide step-by-step answers (e.g., "Chain-of-Thought") is the prominent approach for complex reasoning tasks, where more accurate reasoning chains typically improve ...downstream task performance. Recent literature discusses automatic methods to verify reasoning to evaluate and improve their correctness. However, no fine-grained step-level datasets are available to enable thorough evaluation of such verification methods, hindering progress in this direction. We introduce REVEAL: Reasoning Verification Evaluation, a dataset to benchmark automatic verifiers of complex Chain-of-Thought reasoning in open-domain question-answering settings. REVEAL includes comprehensive labels for the relevance, attribution to evidence passages, and logical correctness of each reasoning step in a language model's answer, across a variety of datasets and state-of-the-art language models. Evaluation on REVEAL shows that verifiers struggle at verifying reasoning chains - in particular, verifying logical correctness and detecting contradictions. Available at https://reveal-dataset.github.io/ .
Periodic behavior in aperiodic multilayers Sharabi, Yonatan; Sheinfux, Hanan Herzig; Eisenstein, Gadi ...
2017 Conference on Lasers and Electro-Optics (CLEO)
Conference Proceeding
We present a family of one-dimensional quasiperiodic crystal which simultaneously display both the fractal bandstructure typical to quasiperiodic structures and properties normally exclusive to ...periodic structures, including Bloch-like modes.
We experimentally demonstrate, for the first time, waveguiding using artificial gauge fields. We use a system of waveguide arrays where the gauge field, arising by tilting the waveguides, affects ...transversal dynamics and generates guided modes.
We show that introducing asymmetric coupling can force all the modes of a waveguide lattice to localize, except for one topologically protected "edge-state" which becomes extended. This mode has real ...eigenvalues and retains topological properties.
We study the interface between two artificial gauge fields in a 2D photonic lattice, and find the analogues of Snell's law and Fresnel coefficients of such interfaces.