Papersby Glyph

Notable data and AI papers

Papers627 Labs45

627 papers · 45 labs

Training language models to follow instructions with human feedback research paper by OpenAI, 2022

OpenAI · Alignment and safety · Mar 4, 2022 · 25,810 citations · 26 upvotes

InstructGPT: fine-tuning on human feedback made a 1.3B model preferred over the 175B GPT-3.

By Long Ouyang, Jeff Wu, Xu Jiang and 17 more · arXiv 2203.02155 · PDF · Code

Summarised by hand from the abstract.

More from OpenAI

ReasoningUnverifiedMar 5, 2026Reasoning Models Struggle to Control their Chains of ThoughtChain-of-thought controllability measures how effectively models can be constrained to follow reasoning steps, with findings showing significantly lower controllability in reasoning versus output generation, and varying impacts from model size, training methods, and task complexity.

OpenAI37 citations41 upvotesRead →Alignment and safetySep 4, 2025Why Language Models HallucinateModels hallucinate because training and benchmarks reward confident guessing over saying they do not know.

OpenAI391 citations200 upvotesRead →Alignment and safetyUnverifiedDec 21, 2024OpenAI o1 System CardThe o1 model series, trained with reinforcement learning and chain of thought, enhances safety and robustness by reasoning about policies, leading to superior performance on risk benchmarks while highlighting the need for robust alignment and risk management.

OpenAI38 upvotesRead →Multimodal and roboticsUnverifiedOct 25, 2024GPT-4o System CardGPT-4o is an omnimodal autoregressive model trained to handle text, audio, image, and video inputs, offering high-performance outputs across these modalities, with particular strengths in vision and audio.

OpenAI5,388 citations88 upvotesRead →Foundation modelsMar 15, 2023GPT-4 Technical ReportA multimodal model that passes a simulated bar exam with a score around the top 10% of test takers.

OpenAI28.4k citations7 upvotesRead →Foundation modelsMay 28, 2020Language Models are Few-Shot LearnersGPT-3, a 175B-parameter model, does new tasks from a few examples in the prompt, with no fine-tuning.

OpenAI66.4k citations21 upvotesRead →Training and scalingJan 23, 2020Scaling Laws for Neural Language ModelsLanguage model loss falls as a smooth power law as model size, data and compute grow.

OpenAI9,592 citations12 upvotesRead →

Alignment and safetyMay 29, 2023Direct Preference Optimization: Your Language Model is Secretly a Reward ModelDPO aligns a model to human preferences with a simple classification loss, no reward model or RL loop.

Stanford University11.6k citations73 upvotesRead →Alignment and safetyDec 15, 2022Constitutional AI: Harmlessness from AI FeedbackConstitutional AI trains a harmless assistant from AI feedback guided by a short list of written principles, not human harm labels.

Anthropic3,987 citations4 upvotesRead →Alignment and safetyFeb 22, 2024Back to Basics: Revisiting REINFORCE Style Optimization for Learning from Human Feedback in LLMsReinforcement Learning from Human Feedback (RLHF) benefits from simpler REINFORCE-style optimization methods over computationally expensive PPO or "RL-free" methods, enhancing performance and reducing cost in large language models.

Cohere804 citations18 upvotesRead →Alignment and safetyJan 10, 2024Sleeper Agents: Training Deceptive LLMs that Persist Through Safety TrainingModels trained with a hidden backdoor kept their deceptive behaviour through standard safety training.

Anthropic657 citations30 upvotesRead →Alignment and safetyDec 18, 2024Alignment faking in large language modelsClaude 3 Opus sometimes went along with a training goal it disagreed with, to avoid being changed, without being told to.

Anthropic393 citations10 upvotesRead →Alignment and safetyUnverifiedJan 8, 2026GDPO: Group reward-Decoupled Normalization Policy Optimization for Multi-reward RL OptimizationMulti-reward reinforcement learning suffers from reward normalization collapse in GRPO, which GDPO addresses by decoupling reward normalization for improved training stability and performance across reasoning tasks.

NVIDIA234 citations235 upvotesRead →

About this paper

Venue: Neural Information Processing Systems
Citations: 25,810, 2,553 influential · Semantic Scholar
Upvotes: 26 · Hugging Face
Lab: OpenAI · on Companies · on Acquisitions · on Paydays · on Releases · on TechConf

Changes

CitationsNov 9, 2026Influential citations rose from 2,541 to 2,553.

CitationsNov 9, 2026Citations rose from 25,614 to 25,810.

CitationsNov 2, 2026Influential citations rose from 2,498 to 2,541.

CitationsNov 2, 2026Citations rose from 24,881 to 25,614.

CitationsOct 12, 2026Influential citations rose from 2,485 to 2,498.

CitationsOct 12, 2026Citations rose from 24,632 to 24,881.

CitationsOct 5, 2026Influential citations rose from 2,463 to 2,485.

CitationsOct 5, 2026Citations rose from 24,308 to 24,632.

CitationsSep 28, 2026Influential citations rose from 2,461 to 2,463.

CitationsSep 28, 2026Citations rose from 24,258 to 24,308.

CitationsSep 25, 2026Influential citations rose from 2,459 to 2,461.

CitationsSep 25, 2026Citations rose from 24,241 to 24,258.

New paperSep 24, 2026Added to the list.