ABCDEFGHIJKLMNOPQRSTUVWXYZAAABAC
1
TypeDateTitleImageLinkContent
2
Paper1/09/2023RLAIF: Scaling Reinforcement Learning from Human Feedback with AI Feedbackhttps://arxiv.org/abs/2309.00267RLAIF's AI-driven feedback matches human feedback in aligning LLMs, with both outperforming the baseline in 70% of tests, offering a scalable alternative to RLHF.
3
Model04/09/2023TinyLlama
https://github.com/jzhang38/TinyLlama/raw/main/.github/TinyLlama_logo.png
https://github.com/jzhang38/TinyLlamaA 1.1B-parameter Llama 2 clone pretrained on 3T tokens in 90 days with 16 GPUs—compact power for projects needing low compute footprint.
4
Paper05/09/2023Cognitive Architecture for Language Agentshttps://arxiv.org/abs/2309.02427Drawing from symbolic AI's agent design, they introduce the CoALA framework, blending LLMs with cognitive architectures for systematic development of advanced language agents.
5
Model6/09/2023Falcon 180B
https://encrypted-tbn2.gstatic.com/images?q=tbn:ANd9GcR3EnofDsPdbMmBW47p610_zQzX8kNUzzUZx0DnjaPNkIlWcKw6
https://huggingface.co/blog/falcon-180bAbu Dhabi's Tech Innovation Institute introduces Falcon 180B, a open-use top-ranking language model rivaling GPT-4 and Google's PaLM 2 Large.
6
Paper06/09/2023GPT Can Solve Mathematical Problems Without a Calculatorhttps://arxiv.org/abs/2309.03241A 2 billion-parameter model outperforms GPT-4 in multi-digit arithmetic if trained with sufficient data and matches GPT-4 in Chinese math problems when fine-tuned.
7
Paper7/09/2023Large Language Models as Optimizershttps://arxiv.org/abs/2309.03409Optimization by PROmpting (OPRO) method uses LLMs to optimize tasks described in natural language, surpassing human prompts by up to 50%.
8
Model7/09/2023FLM-101Bhttps://huggingface.co/CofeAI/FLM-101B101B-parameter LLM trained under $100K budget matches GPT-3 performance with innovative IQ-evaluation.
9
News8/09/2023TensorRT-LLM
NVIDIA TensorRT-LLM Supercharges Large Language Model Inference on NVIDIA H100 GPUs | NVIDIA Technical Blog
NVIDIA unveils TensorRT-LLM for optimizing large language models, supports popular LLMs and it will be integrated into the NVIDIA NeMo framework.
10
Paper8/09/2023From Sparse to Dense: GPT-4 Summarization with Chain of Density Promptinghttps://arxiv.org/abs/2309.04269The "Chain of Density" (CoD) prompt enables GPT-4 to produce denser summaries preferred by humans, balancing informativeness and readability.
11
Model9/09/2023Baichuan 2https://huggingface.co/baichuan-incBaichuan 2, a large-scale multilingual language model with up to 13 billion parameters, outperforms similar open-source models on key benchmarks and excels in domains like medicine and law.
12
Paper11/09/2023NExT-GPT: Any-to-Any Multimodal LLMhttps://arxiv.org/abs/2309.05519NExT-GPT enables any-to-any content generation across text, images, videos, and audio with minimal parameter tuning of an only-text LLM, advancing human-like AI interaction.
13
Paper11/09/2023Textbooks Are All You Need II: phi-1.5 technical reporthttps://arxiv.org/abs/2309.05463The study unveils phi-1.5, a 1.3 billion parameter model that matches larger models in performance, using "textbook quality" data for improved reasoning and is open-sourced for further exploration.
14
Paper12/09/2023A Survey of Hallucination in Large Foundation Modelshttps://arxiv.org/abs/2309.05922This survey dives deep into the concept of 'hallucination' in LLMs, classifying its types, establishing evaluation standards, reviewing mitigation strategies, and charting the path for future research.
15
News13/09/2023Open Interpreterhttps://github.com/KillianLucas/open-interpreter/A permisively licensed alternative to OpenAI's CodeInterpreter that allows diferent LLMs to run locally and use natural language to run code.
16
Paper15/09/2023Connecting Large Language Models with Evolutionary Algorithms Yields Powerful Prompt Optimizershttps://arxiv.org/abs/2309.08532The paper introduces "EvoPrompt", a framework using evolutionary algorithms to optimize prompts for LLMs, yielding significant improvements over human-crafted and other automatic methods.
17
Paper17/09/2023Contrastive Decoding Improves Reasoning in Large Language Modelshttps://arxiv.org/abs/2309.09117Contrastive Decoding improves over existing methods by preventing some abstract reasoning errors, as well as by avoiding simpler modes such as copying sections of the input during chain-of-thought.
18
Paper18/09/2023Adapting Large Language Models via Reading Comprehensionhttps://arxiv.org/abs/2309.09530By transforming raw corpora into reading comprehension texts, domain-specific training for LLMs is improved, achieving competitive performance across multiple domains, even rivaling larger specialized models.
19
Paper19/09/2023LMDX: Language Model-based Document Information Extraction and Localizationhttps://arxiv.org/abs/2309.10952LMDX is a novel method to adapt LLMs for extracting and localizing information from visually rich documents, surpassing previous benchmarks and ensuring data efficiency.
20
Paper19/09/2023Language Modeling Is Compressionhttps://arxiv.org/abs/2309.10668LLMs, due to their predictive prowess, can be powerful compressors; Chinchilla 70B outperforms domain-specific compressors, highlighting the interconnectedness of prediction and compression.
21
Paper20/09/2023Chain-of-Verification Reduces Hallucination in Large Language Modelshttps://arxiv.org/abs/2309.11495Chain-of-Verification makes the model fact check its response iteratively reducing the amount of plausible yet incorrect factual information that it produces.
22
Paper20/09/2023A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Modelshttps://arxiv.org/abs/2309.11674Using a two-stage fine-tuning process on both monolingual and high-quality parallel data significantly improves translation performance of moderate-sized LLMs, outperforming even larger models.
23
Model20/09/2023BTLM-3B-8Khttps://arxiv.org/abs/2309.11568BTLM-3B-8K sets a new standard, outperforming other 3B models and even rivaling some 7B models; optimized for mobile and edge devices, it's available under Apache 2.0.
24
Paper21/09/2023LongLoRA: Efficient Fine-tuning of Long-Context Large Language Modelshttps://arxiv.org/abs/2309.12307An efficient fine-tuning approach that extends the context sizes of pre-trained LLMs reducing the computational costs of training for longer context lengths.
25
Paper23/09/2023Exploring Large Language Models' Cognitive Moral Development through Defining Issues Testhttps://arxiv.org/abs/2309.13356This study builds a bridge between the fields of psychology and AI evaluating LLMs' ethical reasoning using psychology principles and the Psychometric Assessment Tool-Defining Issues Test.
26
Model25/09/2023Qwen
https://github.com/QwenLM/Qwen/blob/main/assets/logo.jpg
https://github.com/QwenLM/Qwen7B and14B parameter models that improve the current state of the art models of similar size. Allows for context lenghts up to 8192.
27
Paper26/09/2023QA-LoRA: Quantization-Aware Low-Rank Adaptation of Large Language Modelshttps://arxiv.org/abs/2309.14717The paper creates QA-LoRA which introduces quantization to the fine-tuning process to get less acccuracy loss while reducing training time and memory costs.
28
Model27/09/2023Mistral 7B
https://techcrunch.com/wp-content/uploads/2023/09/mistral-7b-v0.1.jpg?w=990&crop=1
https://huggingface.co/mistralai/Mistral-7B-v0.1One of the best performing 7B parameter models to date, outperforming bigger models like Llama 2 13B on all benchmarks. Released under Apache 2.0 license so it can be used without restrictions.
29
Paper27/09/2023Effective Long-Context Scaling of Foundation Modelshttps://arxiv.org/abs/2309.16039Meta introduces Llama-2-Long with up to 32,768 tokens context window, outperforming previous models on long-context tasks through a unique pretraining approach.
30
News28/09/2023Amazon Bedrock
https://upload.wikimedia.org/wikipedia/commons/thumb/9/93/Amazon_Web_Services_Logo.svg/2560px-Amazon_Web_Services_Logo.svg.png
https://aws.amazon.com/bedrock/Amazon Bedrock is a fully managed service that offers a choice of high-performing foundation models from leading AI companies, simplifying the development of generative AI applications.
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100