Daniel Fleischer

Daniel Fleischer, Moshe Berchansky, Moshe Wasserblat et al. · microsoft-research

Implementing Retrieval-Augmented Generation (RAG) systems is inherently complex, requiring deep understanding of data, use cases, and intricate design decisions. Additionally, evaluating these systems presents significant challenges, necessitating assessment of both retrieval accuracy and generative quality through a multi-faceted approach. We introduce RAG Foundry, an open-source framework for augmenting large language models for RAG use cases. RAG Foundry integrates data creation, training, inference and evaluation into a single workflow, facilitating the creation of data-augmented datasets for training and evaluating large language models in RAG settings. This integration enables rapid prototyping and experimentation with various RAG techniques, allowing users to easily generate datasets and train RAG models using internal or specialized knowledge sources. We demonstrate the framework effectiveness by augmenting and fine-tuning Llama-3 and Phi-3 models with diverse RAG configurations, showcasing consistent improvements across three knowledge-intensive datasets. Code is released as open-source in https://github.com/IntelLabs/RAGFoundry.

Daniel Fleischer, Moshe Berchansky, Gad Markovits et al.

In the rapidly evolving field of Natural Language Processing, Large Language Models (LLMs) are tasked with increasingly complex reasoning challenges. Traditional methods like chain-of-thought prompting have shown promise but often fall short in fully leveraging a model's reasoning capabilities. This paper introduces SQuARE (Sequential Question Answering Reasoning Engine), a novel prompting technique designed to improve reasoning through a self-interrogation paradigm. Building upon CoT frameworks, SQuARE prompts models to generate and resolve multiple auxiliary questions before tackling the main query, promoting a more thorough exploration of various aspects of a topic. Our expansive evaluations, conducted with Llama 3 and GPT-4o models across multiple question-answering datasets, demonstrate that SQuARE significantly surpasses traditional CoT prompts and existing rephrase-and-respond methods. By systematically decomposing queries, SQuARE advances LLM capabilities in reasoning tasks. The code is publicly available at https://github.com/IntelLabs/RAG-FiT/tree/square.

Moshe Berchansky, Daniel Fleischer, Moshe Wasserblat et al. · microsoft-research

State-of-the-art performance in QA tasks is currently achieved by systems employing Large Language Models (LLMs), however these models tend to hallucinate information in their responses. One approach focuses on enhancing the generation process by incorporating attribution from the given input to the output. However, the challenge of identifying appropriate attributions and verifying their accuracy against a source is a complex task that requires significant improvements in assessing such systems. We introduce an attribution-oriented Chain-of-Thought reasoning method to enhance the accuracy of attributions. This approach focuses the reasoning process on generating an attribution-centric output. Evaluations on two context-enhanced question-answering datasets using GPT-4 demonstrate improved accuracy and correctness of attributions. In addition, the combination of our method with finetuning enhances the response and attribution accuracy of two smaller LLMs, showing their potential to outperform GPT-4 in some cases.

Marcin Spoczynski, Daniel Fleischer, Moshe Berchansky et al.

Porting deep learning algorithms to new hardware accelerators requires developers to repeatedly apply the same low-level optimizations -- quantization, memory access coalescing, tile size tuning, and architecture-specific workarounds -- to every Triton kernel in their code-base. This manual, repetitive effort is a major bottleneck: each kernel demands the same cycle of trial-and-error profiling against hardware constraints that vary across devices, yet the underlying optimization patterns remain largely consistent. We present Xe-Forge, a multi-stage LLM-powered pipeline that automates this process for Intel GPU. Given a functionally correct Triton kernel, the system applies up to nine optimization stages -- from algorithmic restructuring and operator fusion through block pointer modernization, GPU-specific tuning, and open-ended discovery -- each driven by a Chain-of-Verification-and-Refinement (CoVeR) agent that generates candidates, validates them on real hardware, and iterates on failures. A curated knowledge base encodes Intel GPU constraints (power-of-two warp counts, GRF modes, SLM sizing) that are absent from LLM training data, keeping the model within architecturally valid bounds. We evaluate Xe-Forge on 97 Level-2 KernelBench kernels and Flash Attention on the Intel Arc Pro B70, achieving a 1.17x geometric mean speedup over PyTorch eager with 67% of kernels improving, nine kernels exceeding 5x (up to 82x), and 2--13.3x speedups on Flash Attention across all tested configurations without regression -- demonstrating that structured domain knowledge with hardware-in-the-loop verification can systematically eliminate the repetitive porting effort that currently gates algorithm deployment on new accelerators.

Alon Rozental, Zohar Kelrich, Daniel Fleischer

This paper describes a language representation model which combines the Bidirectional Encoder Representations from Transformers (BERT) learning mechanism described in Devlin et al. (2018) with a generalization of the Universal Transformer model described in Dehghani et al. (2018). We further improve this model by adding a latent variable that represents the persona and topics of interests of the writer for each training example. We also describe a simple method to improve the usefulness of our language representation for solving problems in a specific domain at the expense of its ability to generalize to other fields. Finally, we release a pre-trained language representation model for social texts that was trained on 100 million tweets.

Alon Rozental, Daniel Fleischer, Zohar Kelrich

This paper describes the system developed at Amobee for the WASSA 2018 implicit emotions shared task (IEST). The goal of this task was to predict the emotion expressed by missing words in tweets without an explicit mention of those words. We developed an ensemble system consisting of language models together with LSTM-based networks containing a CNN attention mechanism. Our approach represents a novel use of language models (specifically trained on a large Twitter dataset) to predict and classify emotions. Our system reached 1st place with a macro $\text{F}_1$ score of 0.7145.

Alon Rozental, Daniel Fleischer

This paper describes the participation of Amobee in the shared sentiment analysis task at SemEval 2018. We participated in all the English sub-tasks and the Spanish valence tasks. Our system consists of three parts: training task-specific word embeddings, training a model consisting of gated-recurrent-units (GRU) with a convolution neural network (CNN) attention mechanism and training stacking-based ensembles for each of the sub-tasks. Our algorithm reached 3rd and 1st places in the valence ordinal classification sub-tasks in English and Spanish, respectively.

Alon Rozental, Daniel Fleischer

This paper describes the Amobee sentiment analysis system, adapted to compete in SemEval 2017 task 4. The system consists of two parts: a supervised training of RNN models based on a Twitter sentiment treebank, and the use of feedforward NN, Naive Bayes and logistic regression classifiers to produce predictions for the different sub-tasks. The algorithm reached the 3rd place on the 5-label classification task (sub-task C).