SOTAVerified

Instruction Following

Instruction following is the basic task of the model. This task is dedicated to evaluating the ability of the large model to follow human instructions. It is hoped that the model can generate controllable and safe answers.

Papers

Recently AddedMost HypedMost ActiveNeeds VerificationMost Verified

Showing 751–775 of 1135 papers

TitleStatusHype
On Instruction-Finetuning Neural Machine Translation Models—0
RevisEval: Improving LLM-as-a-Judge via Response-Adapted References—0
Superficial Safety Alignment Hypothesis—0
SFTMix: Elevating Language Model Instruction Tuning with Mixup Recipe—0
CS4: Measuring the Creativity of Large Language Models Automatically by Controlling the Number of Story-Writing ConstraintsCode0
TICKing All the Boxes: Generated Checklists Improve LLM Evaluation and Generation—0
CommonIT: Commonality-Aware Instruction Tuning for Large Language Models via Data PartitionsCode0
SAG: Style-Aligned Article Generation via Model Collaboration—0
Self-Powered LLM Modality Expansion for Large Speech-Text ModelsCode0
Video Instruction Tuning With Synthetic Data—0
LoGra-Med: Long Context Multi-Graph Alignment for Medical Vision-Language Model—0
LLaVA-Critic: Learning to Evaluate Multimodal Models—0
Better Instruction-Following Through Minimum Bayes Risk—0
The Perfect Blend: Redefining RLHF with Mixture of Judges—0
Revisiting the Superficial Alignment Hypothesis—0
Align^2LLaVA: Cascaded Human and Large Language Model Preference Alignment for Multi-modal Instruction CurationCode0
MMMT-IF: A Challenging Multimodal Multi-Turn Instruction Following Benchmark—0
Inference-Time Language Model Alignment via Integrated Value Guidance—0
EAGLE: Towards Efficient Arbitrary Referring Visual Prompts Comprehension for Multimodal Large Language Models—0
Mitigating the Bias of Large Language Model EvaluationCode0
FMDLlama: Financial Misinformation Detection based on Large Language ModelsCode0
Style Outweighs Substance: Failure Modes of LLM Judges in Alignment BenchmarkingCode0
Repairs in a Block World: A New Benchmark for Handling User Corrections with Multi-Modal Language ModelsCode0
Eliciting Instruction-tuned Code Language Models' Capabilities to Utilize Auxiliary Function for Code Generation—0
CamelEval: Advancing Culturally Aligned Arabic Language Models and Benchmarks—0
Show:102550
← PrevPage 31 of 46Next →

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1AutoIF (Llama3 70B)Inst-level loose-accuracy90.4—Unverified
2AutoIF (Qwen2 72B)Inst-level loose-accuracy88—Unverified
3GPT-4Inst-level loose-accuracy85.37—Unverified
4PaLM 2 SInst-level loose-accuracy59.11—Unverified