MemBench

MemBench A Benchmark for Memorability Feedback Training and Evaluation To learn how to use it, visit the official codebase laitifranz/MemCoach. Overview MemBench is a benchmark dataset introduced alongside MemCoach, a training-free method for providing actionable memorability feedback using Multimodal Large Language Mo

Papers using MemBench (7)

Cheers: Decoupling Patch Details from Semantic Representations Enables Unified Multimodal Comprehension and Generation2026STEP3-VL-10B Technical Report2026How to Take a Memorable Picture? Empowering Users with Actionable Feedback2026EvolveMem:Self-Evolving Memory Architecture via AutoResearch for LLM Agents2026Anticipate and Learn: Unleashing Idle-Time Compute in Proactive Agents2026Open-Qwen2VL: Compute-Efficient Pre-Training of Fully-Open Multimodal LLMs on Academic Resources2025Enhancing Instruction-Following Capability of Visual-Language Models by Reducing Image Redundancy2024