V
主页
Branch-Solve-Merge Improves Large Language Model Evaluation and Generation
发布人
论文简述:在这篇名为Branch-Solve-Merge Improves Large Language Model Evaluation and Generation的论文中,作者提出了一种名为Branch-Solve-Merge(BSM)的方法来解决复杂自然语言任务。这种方法包括三个模块:分支、解决和合并模块,这些模块通过特定提示与基础LLM进行参数化。这三个模块计划将任务分解为多个并行子任务,独立解决它们,并将子任务的解决方案融合在一起。作者将这些方法应用于LLM响应评估和受约束文本生成任务中,并使用Vicuna、LLama-2-chat和GPT-4等多个LLM进行评估其有效性。BSM通过提高人类与LLM之间的共识达至26%,减少长度和成对位置偏倚高达50%,使LLama-2-chat在大多数领域上能与或超越GPT-4的表现。在受约束故事生成任务中,BSM提高了故事的连贯性,同时还能提升约束满足度12%。总之,这篇论文提出了一种名为Branch-Solve-Merge的方法来提高大型语言模型的评估和生成能力。通过将任务分解为多个子任务并独立解决它们,这种方法可以提高LLM的性能并在复杂自然语言任务中取得更好的结果。 论文链接: https://arxiv.org/pdf/2310.15123
打开封面
下载高清视频
观看高清视频
视频下载器
FlashDecoding++: Faster Large Language Model Inference on GPUs
Amortizing intractable inference in large language models
MusicAgent: An AI Agent for Music Understanding and Generation with Large Langua
Sheared LLaMA: Accelerating Language Model Pre-training via Structured Pruning
BitNet: Scaling 1-bit Transformers for Large Language Models
LoftQ: LoRA-Fine-Tuning-Aware Quantization for Large Language Models
NEWTON: Are Large Language Models Capable of Physical Reasoning?
Can Large Language Models be Good Path Planners? A Benchmark and Investigation o
Are Large Language Models Post Hoc Explainers?
FreshLLMs: Refreshing Large Language Models with Search Engine Augmentation
[ICLR2024] Skeleton-of-Thought: Prompting Large Language Models for Efficient Pa
Interactive Task Planning with Language Models
ShortGPT: Layers in Large Language Models are More Redundant Than You Expect
Compressing Context to Enhance Inference Efficiency of Large Language Models
Toward Joint Language Modeling for Speech Units and Text
LLM Comparator: Visual Analytics for Side-by-Side Evaluation of Large Language M
[MLSys2024] FlashDecoding++: Faster Large Language Model Inference with Asynchro
databricks《大语言模型:从生产到应用|Large Language Models: Application through Production》
EvalCrafter: Benchmarking and Evaluating Large Video Generation Models
ChatCoder: Chat-based Refine Requirement Improves LLMs' Code Generation
The Era of 1-bit LLMs: All Large Language Models are in 1.58 Bits
How Do Large Language Models Capture the Ever-changing World Knowledge? A Review
Survey on Factuality in Large Language Models: Knowledge, Retrieval and Domain-S
The Consensus Game: Language Model Generation via Equilibrium Search
Ranking LLM-Generated Loop Invariants for Program Verification
斯坦福大学《自然语言处理2023|CS224n: Natural Language Processing》中英字幕
杜克大学《本地大语言模型的基础|Foundations of Local Large Language models》中英字幕
Creative Robot Tool Use with Large Language Models
Llemma: An Open Language Model For Mathematics
CLEX: Continuous Length Extrapolation for Large Language Models
Exponentially Faster Language Modelling
Language Models can be Logical Solvers
Aligner: 一种基于残差思想的大语言模型对齐器(组会分享)
GraphLLM: Boosting Graph Reasoning Ability of Large Language Model
Large Language Models Cannot Self-Correct Reasoning Yet
The FinBen: An Holistic Financial Benchmark for Large Language Models
DPM-Solver-v3: Improved Diffusion ODE Solver with Empirical Model Statistics
[FPGA2024] FlightLLM: Efficient Large Language Model Inference with a Complete M
Let's Synthesize Step by Step: Iterative Dataset Synthesis with Large Language M
LAMP: Learn A Motion Pattern for Few-Shot-Based Video Generation