Market/News Routing Model (MuRIL, Hindi/Punjabi)

MuRIL fine-tuned for native Hindi/Punjabi market/news category classification (agri_commodity/real_estate/equities_markets/ policy_regulatory/other). AutoScientist Challenge Part 2, Market Analysis & News category.

Results

Accuracy 62.5%, Macro-F1 0.320 (held-out 20% split, 80 rows).

Class Precision Recall F1 Support
other 0.81 0.59 0.68 51
policy_regulatory 0.47 0.83 0.60 24
agri_commodity 0.00 0.00 0.00 2
real_estate 0.00 0.00 0.00 3
equities_markets 0.00 0.00 0.00 0

Known limitation, disclosed honestly: category distribution is real IndicCorpV2 web-crawl data, not curated for balance. The three finance-specific categories have too few training examples (2-14 rows) for the model to learn reliably; it performs respectably on the two well-represented classes. See the source project's PART2_SUBMISSION.md for the full data-sourcing writeup.

Downloads last month
17
Safetensors
Model size
0.2B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for tojpaj/marketnews-routing-model

Finetuned
(71)
this model

Dataset used to train tojpaj/marketnews-routing-model

Space using tojpaj/marketnews-routing-model 1