Andrew Drozdov mrdrozdov

Deconstruct 2017 Bibliography

Here are all of the resources mentioned by Deconstruct 2017 speakers, along with who recommended what. Please post a comment if I missed something or have an error!

DC 2017 Speakers' Choice Gold Medalist

Seeing Like a State by James Scott

Books

Public Opinion by Walter Lippmann (Evan Czaplicki)
A Pattern Language by Christopher Alexander (Brian Marick)
Domain Driven Design by Eric Evans (Brian Marick)

Reinforcement Learning for Language Models

Yoav Goldberg, April 2023.

Why RL?

With the release of the ChatGPT model and followup large language models (LLMs), there was a lot of discussion of the importance of "RLHF training", that is, "reinforcement learning from human feedback". I was puzzled for a while as to why RL (Reinforcement Learning) is better than learning from demonstrations (a.k.a supervised learning) for training language models. Shouldn't learning from demonstrations (or, in language model terminology "instruction fine tuning", learning to immitate human written answers) be sufficient? I came up with a theoretical argument that was somewhat convincing. But I came to realize there is an additional argumment which not only supports the case of RL training, but also requires it, in particular for models like ChatGPT. This additional argument is spelled out in (the first half of) a talk by John Schulman from OpenAI. This post pretty much

	from __future__ import print_function
	import torch
	import torch.nn as nn
	import torch.nn.functional as F
	from torch.autograd import Variable

	def sample_gumbel(shape, eps=1e-20):
	U = torch.rand(shape).cuda()
	return -Variable(torch.log(-torch.log(U + eps) + eps))

	#!/usr/bin/env python3

	"""
	Example of a generic Mixout implementation. (Lee et al., 2019).
	https://arxiv.org/abs/1909.11299

	Implementation by Stephen Roller (https://stephenroller.com).

	Updated 2020-02-10 to include 1/(1 - p) correction term. Thanks to
	Cheolhyoung Lee for making this correction.

	import re


	strings = set()

	with open('texts.txt', mode='r', encoding='utf-8') as file_input:
	content = file_input.read()
	regExStr = r'^\s\d string title = "(?:tooltip\d+\|Name\|Dialogue Text)"\n\s\d string value = "((?!(START\|input\|\w+\(\)\|\!\(\w+\(\)\))).+)"$'
	compiled = re.compile(regExStr, re.MULTILINE)
	matched = compiled.finditer(content)