reflexion
reflexion copied to clipboard
noahshinn
[NeurIPS 2023] Reflexion: Language Agents with Verbal Reinforcement Learning
For example, like this.This is from my trial_0.log： > go to cabinet 3 On the cabinet 3, you see a cup 1. > think: I found a cup in cabinet...
I was running the script https://github.com/noahshinn/reflexion/blob/main/programming_runs/run_reflexion_codellama_multi.sh with CodeLLaMA model, simply change the `codellama` to `codellama-7b` ```bash CUDA_VISIBLE_DEVICES=$1 python main.py \ --run_name "reflexion_codellama_$1" \ --root_dir "root" \ --dataset_path ./benchmarks/humaneval-py.jsonl \ --strategy...
hi there, I am a bit confused about the reflexion for webshop. in code here, line 245, https://github.com/noahshinn/reflexion/blob/main/webshop_runs/webshop_trial.py the llm actor only intakes the `base_prompt + prompt`, which is the...
I was trying the script with CodeLLaMA Sometimes I found the script just got killed, without showing any error. Any intuition? 
I tried Llama-2-7b-chat-hf model on the programming tasks using the prompt for CodeLlama model. The dataset I use is `mbpp-py.jsonl`. But it seems like the prompt doesn't suit the Llama-2-7b-chat-hf...
Hi, Thanks for the great work. Unfortunately, we are unable to reproduce your results for ReAct / Reflexion on Alfworld. E.g. Env0 & Env1 are successful for you, however, we...
I'd like to run your code as it was implemented for the original paper. Would it be possible to add a link in the README that points to the specific...
It seems like the `./run_simple.sh` and `./run_reflexion.sh` scripts run through all the environments one by one. It is not clear, however, where in the code the particular environment is selected....
Hi Noah, I'm reproducing your work, generally I view reflexion as some kind of in-context few-shot sft/rl, which requires supervised signals (either from environment or label) . However, in your...
