AI Papers of the Week
Every paper worth reading in AI, hand-picked one week at a time.
Unifying LLMs & Knowledge Graphs
A roadmap for combining LLMs with knowledge graphs for stronger reasoning.
Fine-Grained RLHF
Trains LMs with segment-level human feedback rather than whole-response preferences.
Hierarchical Vision Transformer (Hiera)
Pretrains ViTs with MAE while removing unnecessary multi-stage complexity.
Imitating Reasoning Process of Larger LLMs (Orca)
Microsoft's 13B model that imitates GPT-4's reasoning traces.
Fine-Tuning Language Models with Just Forward Passes (MeZO)
A memory-efficient zeroth-order optimizer for LLM fine-tuning.
MERT
An acoustic music understanding model with large-scale self-supervised training.
Bytes Are All You Need
Performs classification directly on file bytes without decoding.
Direct Preference Optimization (DPO)
Rafailov et al.'s simpler alternative to RLHF that rivals full RL-based alignment.
SQL-PaLM
An LLM-based Text-to-SQL system built on PaLM-2.
CodeTF
An open-source Transformer library for state-of-the-art code LLMs.
QLoRA
Tim Dettmers' breakthrough technique enabling 65B LLM fine-tuning on a single 48GB GPU.
LIMA
Meta's 65B LLaMA fine-tuned on just 1,000 curated examples - showing alignment needs less data than believed.
Gorilla
A fine-tuned LLaMA-based model that surpasses GPT-4 on API call generation.
The False Promise of Imitating Proprietary LLMs
Berkeley's critical analysis of open-source imitation of proprietary LLMs.
Sophia
A simple, scalable second-order optimizer with negligible per-step overhead.
Reinventing RNNs for the Transformer Era (RWKV)
Combines parallelizable training of Transformers with efficient RNN inference.
DoReMi
Optimizes data mixtures for faster language model pretraining.
CodeT5+
An open code LLM family for code understanding and generation.
Symbol tuning
Fine-tunes LMs on in-context input-label pairs with natural-language labels replaced by arbitrary symbols.
Incidental Bilingualism in PaLM's Translation Capability
Explores where PaLM's translation ability actually comes from.
InstructBLIP
Visual-language instruction tuning built on BLIP-2.
MultiModal-GPT
A vision-language model for multi-round dialogue fine-tuned from OpenFlamingo.
scGPT
A foundation model for single-cell multi-omics pretrained on 10 million cells.
GPTutor
A ChatGPT-powered VSCode extension for code explanation.