language-agents

A language agent gym with challenging scientific tasks

[NeurIPS 2024] Official implementation for paper "Can Graph Learning Improve Planning in LLM-based Agents?"

Benchmark for LLM-based Agents in Computational Biology

Framework enabling modular interchange of language agents, environments, and optimizers

Optimal stopping agents powered by generative large language models

Evidence-Bound Retrieval (EBR) — accretive memory for AI agents. Amendment records, structural precedence, BM25. A substrate for CoALA's episodic memory layer.

Execution-time action control for ReAct language agents with stage-aware guards, wrong-object prevention, and action redirection in ALFWorld.

Explore the capabilities of LLMs within agentic architectures for Georelating natural disaster events.

Investigating whether a persistent caregiver–child relationship improves knowledge transfer in language agents.

Experiments on generative and cognitive agents from my MSc thesis: purely generative (LLM-based), CoALA-inspired and symbolic (ASTRA) agents compared in a dynamic, partially observable, stochastic environment

twagents is a framework for building and comparing language model agents with different memory architectures in interactive, text-based environments. It is designed to work with TextWorld and focuses on exploring how working, episodic, and semantic memory influence agent behavior, recall, and decision-making.

Cognitive AutoGen (CoAG) extends Microsoft AutoGen with a CoALA-inspired cognitive layer — structured working memory, RAG long-term memory and a 6-phase deliberative cycle