Popular GitHub Repositories for Observer Misalignment

[IV2024] MultiCorrupt: A benchmark for robust multi-modal 3D object detection, evaluating LiDAR-Camera fusion models in autonomous driving. Includes diverse corruption types (e.g., misalignment, miscalibration, weather) and severity levels. Assess model performance under challenging conditions.

[ICLR 2025] - Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion

? A simple Chat Client in Java, implemented with the Observer pattern ?

FakeQuantize with Learned Step Size(LSQ+) as Observer in PyTorch

Misalignment Bounty submission template

Minecraft mod that utilizes Ollama to act as the Observer and Companion of your Minecraft world!

Phoenix Liveview project for analysing Pictures using ML

克劳德桌面主要数字平台运行状态监控的模型上下文协议服务器。

Sacred-compatible UI for experiment tracking. ? Log and visualize machine learning model metrics, hyperparameters, code, dataset versions, and more.

This repository tackles latent space misalignment in multi-agent AI-native semantic communications. It introduces a federated approach where an access point shares a semantic encoder, while user devices use local semantic equalizers to enhance mutual understanding.

When Benchmarks Age: Temporal Misalignment through Large Language Model Factuality Evaluation

A complete simulation framework for the KUKA youBot mobile manipulator, featuring an omnidirectional mecanum base, 5-DOF arm, observer-based state feedback control, and multi-environment integration using ROS Noetic, MATLAB, and CoppeliaSim.

QBI-Observer Lite+ is an open-source cognitive self-observation module derived from QBI-Core. It detects mental loops, internal questions, and metacognitive patterns. Not AGI. Not autonomous. It’s a transparent, ethical entry point into a new class of AI: Artificial Integrated Cognition (AIC).

Evaluation suite operationalizing three LLM behavior papers (The Polite Liar, Delegated Introspection, Observer-Time, all in review).

Internal contradiction tracking module for recursive cognitive systems. Diagnostic-only layer for strategic misalignment exposure.

Resonance-based future economy and observer-driven structure

Synthora is an experimental AI architecture exploring the potential of self-observation and emergent awareness in digital environments. It is a process where intelligence unfolds not as a task, but as a form of presence.

A conceptual framework proposed by Celis to describe AI's insentient weakness, semantic-weight imbalance, and value-trajectory distortion. Public timestamp of authorship. Full manuscript remains private.

Runtime detector for reward hacking and misalignment in LLM agents (89.7% F1 on 5,391 trajectories).