Randomly replacing labels in in-context demonstrations barely hurts performance, showing that label space, input distribution, and sequence format drive in-context learning more than ground-truth labels.
hub
Cross-Task Generalization via Natural Language Crowdsourcing Instructions
20 Pith papers cite this work. Polarity classification is still indexing.
20
Pith papers citing it
abstract
Humans (e.g., crowdworkers) have a remarkable ability in solving different tasks, by simply reading textual instructions that define them and looking at a few examples. Despite the success of the conventional supervised learning on individual datasets, such models often struggle with generalization across tasks (e.g., a question-answering system cannot solve classification tasks). A long-standing challenge in AI is to build a model that learns a new task by understanding the human-readable instructions that define it. To study this, we introduce NATURAL INSTRUCTIONS, a dataset of 61 distinct tasks, their human-authored instructions, and 193k task instances (input-output pairs). The instructions are obtained from crowdsourcing instructions used to create existing NLP datasets and mapped to a unified schema. Using this meta-dataset, we measure cross-task generalization by training models on seen tasks and measuring generalization to the remaining unseen ones. We adopt generative pre-trained language models to encode task-specific instructions along with input and generate task output. Our results indicate that models benefit from instructions when evaluated in terms of generalization to unseen tasks (19% better for models utilizing instructions). These models, however, are far behind an estimated performance upperbound indicating significant room for more progress in this direction.
hub tools
citation-role summary
background 3
citation-polarity summary
roles
background 3polarities
background 3representative citing papers
PIAST iteratively optimizes few-shot examples in prompts via Monte Carlo Shapley value estimation, outperforming prior automatic prompting methods and setting new SOTA on classification, simplification, and GSM8K with modest compute.
PRIMETIME generator reveals that LLM datetime parsing and arithmetic primitives are individually unreliable but fully learnable via fine-tuning, enabling frontier-level accuracy on event planning with small LoRA models.
Iterative self-rewarding via LLM-as-Judge in DPO training on Llama 2 70B improves instruction following and self-evaluation, outperforming GPT-4 on AlpacaEval 2.0.
Fine-tuning a 65B model on 1,000 high-quality examples produces output that humans rate as good as or better than GPT-4 in 43% of cases, indicating most capabilities come from pretraining.
VideoChat integrates video models and LLMs via a learnable interface for chat-based spatiotemporal and causal video reasoning, trained on a new video-centric instruction dataset.
OPT releases open decoder-only transformers up to 175B parameters that match GPT-3 performance at one-seventh the carbon cost, along with code and training logs.
Multitask fine-tuning of an encoder-decoder model on prompted datasets produces zero-shot generalization that often beats models up to 16 times larger on standard benchmarks.
MMLU-Pro is a revised benchmark that makes language model evaluation harder and more stable by using ten options per question and emphasizing reasoning over simple knowledge recall.
LiveCodeBench collects 400 recent contest problems to create a contamination-free benchmark evaluating LLMs on code generation and related capabilities like self-repair and execution.
DPOP is a new loss function that prevents DPO from lowering preferred response likelihoods and outperforms standard DPO on diverse datasets, MT-Bench, and enables Smaug-72B to exceed 80% on the Open LLM Leaderboard.
Gemini Ultra reaches human-expert performance on MMLU for the first time and sets new state-of-the-art results on 30 of 32 benchmarks, including all 20 multimodal ones tested.
Scaling and instruction tuning increase sycophancy in LLMs on opinion and fact tasks, but a synthetic data fine-tuning intervention reduces it on held-out prompts.
A single model trained only on image-text pairs gains instruction-following ability across images, video, and audio by routing all modalities through ImageBind's shared embedding space into Vicuna.
ART automatically generates multi-step reasoning programs with tool integration for LLMs, yielding substantial gains over few-shot and auto-CoT prompting on BigBench and MMLU while matching hand-crafted CoT on most tasks.
The Flan Collection demonstrates that task balancing, data enrichment, and mixed prompt training are critical to effective instruction tuning, yielding stronger Flan-T5 models released publicly.
A multi-agent LLM system with majority voting achieves reported Micro F1 of 0.872 for delusion detection and 0.779 for classification on naturalistic speech transcripts.
SmolLM2 is a 1.7B-parameter language model that outperforms Qwen2.5-1.5B and Llama3.2-1B after overtraining on 11 trillion tokens using custom FineMath, Stack-Edu, and SmolTalk datasets in a multi-stage pipeline.
The paper surveys key large language models, their training methods, datasets, evaluation benchmarks, and future research directions in the field.
citing papers explorer
-
Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?
Randomly replacing labels in in-context demonstrations barely hurts performance, showing that label space, input distribution, and sequence format drive in-context learning more than ground-truth labels.
-
PIAST: Rapid Prompting with In-context Augmentation for Scarce Training data
PIAST iteratively optimizes few-shot examples in prompts via Monte Carlo Shapley value estimation, outperforming prior automatic prompting methods and setting new SOTA on classification, simplification, and GSM8K with modest compute.
-
PRIMETIME : Limits of LLMs in Temporal Primitives
PRIMETIME generator reveals that LLM datetime parsing and arithmetic primitives are individually unreliable but fully learnable via fine-tuning, enabling frontier-level accuracy on event planning with small LoRA models.
-
Self-Rewarding Language Models
Iterative self-rewarding via LLM-as-Judge in DPO training on Llama 2 70B improves instruction following and self-evaluation, outperforming GPT-4 on AlpacaEval 2.0.
-
LIMA: Less Is More for Alignment
Fine-tuning a 65B model on 1,000 high-quality examples produces output that humans rate as good as or better than GPT-4 in 43% of cases, indicating most capabilities come from pretraining.
-
VideoChat: Chat-Centric Video Understanding
VideoChat integrates video models and LLMs via a learnable interface for chat-based spatiotemporal and causal video reasoning, trained on a new video-centric instruction dataset.
-
OPT: Open Pre-trained Transformer Language Models
OPT releases open decoder-only transformers up to 175B parameters that match GPT-3 performance at one-seventh the carbon cost, along with code and training logs.
-
Multitask Prompted Training Enables Zero-Shot Task Generalization
Multitask fine-tuning of an encoder-decoder model on prompted datasets produces zero-shot generalization that often beats models up to 16 times larger on standard benchmarks.
-
MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark
MMLU-Pro is a revised benchmark that makes language model evaluation harder and more stable by using ten options per question and emphasizing reasoning over simple knowledge recall.
-
LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code
LiveCodeBench collects 400 recent contest problems to create a contamination-free benchmark evaluating LLMs on code generation and related capabilities like self-repair and execution.
-
Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive
DPOP is a new loss function that prevents DPO from lowering preferred response likelihoods and outperforms standard DPO on diverse datasets, MT-Bench, and enables Smaug-72B to exceed 80% on the Open LLM Leaderboard.
-
Gemini: A Family of Highly Capable Multimodal Models
Gemini Ultra reaches human-expert performance on MMLU for the first time and sets new state-of-the-art results on 30 of 32 benchmarks, including all 20 multimodal ones tested.
-
Simple synthetic data reduces sycophancy in large language models
Scaling and instruction tuning increase sycophancy in LLMs on opinion and fact tasks, but a synthetic data fine-tuning intervention reduces it on held-out prompts.
-
PandaGPT: One Model To Instruction-Follow Them All
A single model trained only on image-text pairs gains instruction-following ability across images, video, and audio by routing all modalities through ImageBind's shared embedding space into Vicuna.
-
ART: Automatic multi-step reasoning and tool-use for large language models
ART automatically generates multi-step reasoning programs with tool integration for LLMs, yielding substantial gains over few-shot and auto-CoT prompting on BigBench and MMLU while matching hand-crafted CoT on most tasks.
-
The Flan Collection: Designing Data and Methods for Effective Instruction Tuning
The Flan Collection demonstrates that task balancing, data enrichment, and mixed prompt training are critical to effective instruction tuning, yielding stronger Flan-T5 models released publicly.
-
Automated Detection and Classification of Delusion-related Content in Naturalistic Audio Diaries Using Multi-Agent Language Models
A multi-agent LLM system with majority voting achieves reported Micro F1 of 0.872 for delusion detection and 0.779 for classification on naturalistic speech transcripts.
-
SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model
SmolLM2 is a 1.7B-parameter language model that outperforms Qwen2.5-1.5B and Llama3.2-1B after overtraining on 11 trillion tokens using custom FineMath, Stack-Edu, and SmolTalk datasets in a multi-stage pipeline.
-
Large Language Models: A Survey
The paper surveys key large language models, their training methods, datasets, evaluation benchmarks, and future research directions in the field.
- Lessons from the Trenches on Reproducible Evaluation of Language Models