Skip to content

Commit 6a3a231

Browse files
authored
Update README.md
1 parent b0cc474 commit 6a3a231

File tree

1 file changed

+24
-1
lines changed

1 file changed

+24
-1
lines changed

project/moirai-agent/README.md

Lines changed: 24 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1 +1,24 @@
1-
This folder temporarily hosts the scripts to replicate early Moirai-agent results on contextual and non-contextual forecasting tasks.
1+
# Moirai Agent
2+
3+
Moirai agent is an intelligent forecasting framework that blends historical time series data with external contextual signals to deliver more robust and adaptive predictions than traditional numerical-only models. It leverages large language models as orchestrators to integrate heterogeneous information, select forecasting experts, and adjust forecasts based on real-world context. This makes it especially effective in dynamic environments where context matters as much as past values. This repository provides an early research preview and enables replication of benchmark results for two independent yet composable agents.
4+
5+
Moirai agent supports two core capabilities. First, it enables context-aware forecasting by combining numerical history with external contextual inputs to refine lookback windows, detect anomalies, and anticipate future effects. Second, it performs expert selection through an LLM-based selector that chooses the most suitable forecasting model for each task from a pool of state-of-the-art forecasters.
6+
7+
```
8+
MoiraiAgent
9+
├── ctx_forecast
10+
│ └── … # Detailed scripts for context-aware forecasting workflows
11+
├── gift_eval
12+
│ └── … # Scripts focused on replicating gift_eval results with standard numeric forecasting through model-selection
13+
└── README.md # Project introduction (this file)
14+
```
15+
16+
👉 For a deeper dive, see the accompanying blog post: [MoiraiAgent blog](https://www.salesforce.com/blog/moiraiagent)
17+
18+
👉 Standard forecasting results are evaluated on GIFT-Eval, a large-scale benchmark for general time series forecasting: [GIFT-Eval leaderboard](https://huggingface.co/spaces/Salesforce/gift-eval)
19+
20+
👉 Contextual forecasting results are evaluated on a newly curated benchmark that combines the [CIK](https://arxiv.org/abs/2410.18959) benchmark with additional samples generated by us: [GIFT-CTX Dataset](https://huggingface.co/datasets/Salesforce/GIFT-CTX)
21+
22+
This is an open-source research version of Moirai-agent, the code is made available purely for research purposes.
23+
24+

0 commit comments

Comments
 (0)