Card: LOTUS logo banner under the title

README.md

@@ -13,7 +13,9 @@ base_model:
 pipeline_tag: time-series-forecasting
 ---
 
-# 🪷 LOTUS: Learning to Understand Time-series by Learning to Forecast
+# LOTUS: Learning to Understand Time-series by Learning to Forecast
+
+<p align="center"><img src="LOTUS.png" alt="LOTUS" width="420"></p>
 
 **LOTUS** is a time-series language model pretrained under **forecast grounding** — a pretraining
 principle that requires language-aligned time-series representations to *remain predictive of the
@@ -36,6 +38,7 @@ baseline by **3.36%** — and its 270M-parameter model is the top system on the
 reasoning benchmark ahead of backbones 25–50× larger.
 
 Paper: [arXiv:2610.18877](https://arxiv.org/abs/2610.18877) · Code: [DA2I2-SLM/Lotus](https://github.com/DA2I2-SLM/Lotus)
+· Phase-1 (pretraining) checkpoints: [hiepnh137/LangTimeFoundationModel](https://huggingface.co/hiepnh137/LangTimeFoundationModel)
 
 ## Training
 
@@ -51,7 +54,7 @@ explicit answer, which is parsed and compared with the reference. Classification
 reported on HAR, PAMAP2, Sleep-EDF and Time-MQA (MCQ and true/false splits), and generative
 accuracy under normalization on ECG-QA.
 
-## Training data
+## Training data (as specified in the paper)
 
 **Phase 1 — forecast-grounded pretraining.** Four corpora: Monash half-hourly electricity
 demand (Godahewala et al., 2021) and a synthetic multivariate corpus for the forecasting
@@ -126,8 +129,6 @@ from load_lotus import load_lotus
 lotus = load_lotus()          # downloads the checkpoint, picks GPU + dtype automatically
 ```
 
-Current version of LOTUS supports 2 types of time-series input. You can either type in the series, or upload the series in .csv form.
-
 **Case 1 — type the series in** (any list of numbers, raw scale, in time order):
 
 ```python