---
title: Quant Letter No. 21: October 2023, Week 3
url: https://www.ml-quant.com/issues/2023-10-16/
site: ML-Quant (https://www.ml-quant.com)
updated: 2026-09-26
license: Summaries CC BY 4.0; links go to the original sources
index: https://www.ml-quant.com/llms.txt
issue_date: 2023-10-16
---


# Quant Letter No. 21: October 2023, Week 3

Sent 2023-10-16. 19 items.

## Machine learning

### Recently Published

- __[Mistral: Superior Language Model](https://arxiv.org/abs/2310.06825)__: Superior Language Model: Mistral 7B v0.1 is a language model with 7 billion parameters that excels in reasoning, mathematics, and code generation, and has a version specifically designed to follow instructions. (2023-10-10, shares: 200) · https://www.ml-quant.com/papers/arxiv/2310.06825/
- __[Ferret: Spatial Referring in Images](https://arxiv.org/abs/2310.07704)__: Spatial Referring in Images: Ferret is a Multimodal Large Language Model that can understand and locate spatial references in images, outperforming other models in region-based and localization-required multimodal chatting. (2023-10-11, shares: 52) · https://www.ml-quant.com/papers/arxiv/2310.07704/
- __[MemGPT: Extended Context in LLMs](https://arxiv.org/abs/2310.08560)__: Extended Context in LLMs: MemGPT is a system that manages different memory levels, providing extended context within large language models' limited context windows, enhancing document analysis and multi-session chat performance. (2023-10-12, shares: 33) · https://www.ml-quant.com/papers/arxiv/2310.08560/
- __[Im4D: Real-Time View Synthesis](https://arxiv.org/pdf/2310.08585.pdf)__: Real-Time View Synthesis: Im4D is a hybrid scene representation that combines grid-based geometry and multi-view image-based appearance for dynamic view synthesis from multi-view videos, providing high-quality rendering and efficient training. (2023-10-12, shares: 19) · https://www.ml-quant.com/papers/arxiv/2310.08585/
- __[Octopus: Vision-Language Programmer](https://arxiv.org/abs/2310.08588)__: Vision-Language Programmer: Octopus is a vision-language model that can interpret an agent's vision and textual task objectives to generate complex action sequences and executable code, showing improved decision-making in various tasks. (2023-10-12, shares: 14) · https://www.ml-quant.com/papers/arxiv/2310.08588/
- __[Transformers for Reinforcement Learning](https://arxiv.org/abs/2310.08566)__: The article presents a theoretical framework for training large transformer models for in-context reinforcement learning, offering the first quantitative analysis of their capabilities. (2023-10-12, shares: 10) · https://www.ml-quant.com/papers/arxiv/2310.08566/
- __[Co-emergence of Modularity in RNNs](https://arxiv.org/abs/2310.07711)__: The study uses a brain-inspired modular training method in machine learning to improve network performance and neuron clustering in compositional cognitive tasks. (2023-10-11, shares: 10) · https://www.ml-quant.com/papers/arxiv/2310.07711/
- __[Spectral Matrix Estimation for RL](https://arxiv.org/abs/2310.06793)__: The research proposes new reinforcement learning algorithms for matrix estimation problems with low-rank structure, offering improved performance guarantees. (2023-10-10, shares: 8) · https://www.ml-quant.com/papers/arxiv/2310.06793/
- __[Learning Image Encoders from Unlabelled Video](https://arxiv.org/abs/2310.08584)__: The study presents a new self-supervised image pretraining method using continuous videos, showing that a single video can compete with ImageNet for various tasks. (2023-10-12, shares: 3) · https://www.ml-quant.com/papers/arxiv/2310.08584/
- __[Dynamic Novel View Synthesis from Monocular Videos](https://arxiv.org/pdf/2310.08587.pdf)__: The paper provides an analysis framework for rendering scenes from novel viewpoints in a monocular video, suggesting improvements over some scene-specific methods. (2023-10-12, shares: 2) · https://www.ml-quant.com/papers/arxiv/2310.08587/

### Historical Trending

- __[FateZero: Text-based Video Editing](https://arxiv.org/abs/2303.09535)__: Text-based Video Editing: The article introduces FateZero, a new method for editing real-world videos using text, which outperforms previous models in maintaining video structure, motion, and frame consistency. (2023-03-16, shares: 822) · https://www.ml-quant.com/papers/arxiv/2303.09535/
- __[LLM-grounded Diffusion: Enhancing Text-to-Image Models](https://arxiv.org/abs/2305.13655v1)__: Enhancing Text-to-Image Models: The study suggests a two-stage process using a pretrained language model to improve image generation accuracy in diffusion models, enabling multi-round scene specification in various languages. (2023-05-23, shares: 179) · https://www.ml-quant.com/papers/arxiv/2305.13655/
- __[SelfCheckGPT: Hallucination Detection for LLMs](http://dx.doi.org/10.48550/arxiv.2303.08896)__: Hallucination Detection for LLMs: The paper presents SelfCheckGPT, a new approach for fact-checking black-box model responses without an external database, proving its superior ability to detect and rank factual and non-factual sentences. (2023-03-15, shares: 99) · https://www.ml-quant.com/papers/arxiv/2303.08896/
- __[Quasi-Arithmetic Mixtures and Divergence Minimization](https://arxiv.org/abs/2209.07481)__: The research analyzes the 'centroid' property in the context of Markov Chain Monte Carlo methods, exploring the relationship between parametric families, quasi-arithmetic means, and divergence functions. (2022-09-15, shares: 92) · https://www.ml-quant.com/papers/arxiv/2209.07481/
- __[StoryBench: Text-to-Video Model Benchmark](https://arxiv.org/abs/2308.11606)__: Text-to-Video Model Benchmark: StoryBench is a new benchmark for assessing text-to-video models, offering tasks of different levels of difficulty and guidelines for human evaluation of video narratives. (2023-08-22, shares: 62) · https://www.ml-quant.com/papers/arxiv/2308.11606/
- __[Tensor Programs VI: Deep Residual Network Feature Learning](https://arxiv.org/abs/2310.02244)__: Deep Residual Network Feature Learning: The research explores depthwise parametrizations in deep residual networks, pinpointing Depth-$\mu$P as the best parametrization for maximizing feature learning and diversity, but notes its limitations in deeper networks. (2023-10-03, shares: 53) · https://www.ml-quant.com/papers/arxiv/2310.02244/
- __[Soundify: Sound-Video Matching](https://arxiv.org/abs/2112.09726)__: Sound-Video Matching: Soundify is a new system that aids video editors in synchronizing sounds with videos, proven to lessen workload and enhance usability in a human evaluation study. (2021-12-17, shares: 48) · https://www.ml-quant.com/papers/arxiv/2112.09726/
- __[GPT-MolBERTa: Molecular Property Prediction Language Model](https://arxiv.org/abs/2310.03030)__: Molecular Property Prediction Language Model: GPT-MolBERTa, a self-supervised language model that uses textual descriptions of molecules to predict their properties, is introduced, demonstrating high performance on various molecule property benchmarks. (2023-09-20, shares: 38) · https://www.ml-quant.com/papers/arxiv/2310.03030/
- __[SALMON: Minimal Human Supervision Language Model Alignment](https://arxiv.org/pdf/2310.05910.pdf)__: Minimal Human Supervision Language Model Alignment: The paper introduces SALMON, a new method for aligning base language models with minimal human supervision using principle-following reward models, showing its superior performance on multiple benchmark datasets. (2023-10-09, shares: 36) · https://www.ml-quant.com/papers/arxiv/2310.05910/