Home
Softono
p

princeton-nlp

Professional software vendor delivering innovative solutions on the Softono platform. Specialized in both open-source and proprietary software development.

Visit Website
Total Products
4

Software by princeton-nlp

tree-of-thought-llm
Open Source

tree-of-thought-llm

Tree of Thoughts LLM is an open-source Python package providing the official implementation of the Tree of Thoughts framework, as presented in the NeurIPS 2023 paper Deliberate Problem Solving with Large Language Models. Designed to enhance large language model reasoning, this tool enables systematic problem-solving by exploring multiple reasoning paths in a tree-like structure rather than relying on linear sequences. It supports diverse search strategies including Breadth-First Search and Depth-First Search, along with flexible state evaluation methods such as value estimation and voting. The library includes pre-configured templates for complex tasks like the Game of 24, creative writing generation, and crossword puzzle solving. Users can easily integrate their own tasks by defining custom thought generation and evaluation logic. The codebase is compatible with Python 3.7 and above, released under the MIT license, and supports various backends including OpenAI GPT models. Its architecture allows researchers

Education & Learning ML Frameworks
6K Github Stars
SimCSE
Open Source

SimCSE

SimCSE is a state-of-the-art framework for learning sentence embeddings using simple contrastive learning techniques, as presented in the EMNLP 2021 paper Simple Contrastive Learning of Sentence Embeddings. The software offers two primary approaches: an unsupervised version that treats an input sentence and its dropout-noised version as positive pairs, and a supervised version that leverages Natural Language Inference datasets by using entailment pairs as positives and contradiction pairs as hard negatives. Designed for high efficiency and ease of use, SimCSE is available as a Python package installable via PyPI or from source. It enables users to encode sentences into dense vector representations, compute cosine similarities between sentence groups, and perform semantic search with optional integration of the Faiss library for scalable index building. The suite includes pre-trained models compatible with Hugging Face Transformers for immediate deployment in tasks such as paraphrase identification, semantic t

ML Frameworks Vector Databases
3.6K Github Stars
PURE
Open Source

PURE

Pure is a free social networking App. it is simple, reliable and it makes it easy to keep in touch with your friends and family. Pure works across mobile devices even on slow internet connections.

Communication AI Tools ML Frameworks
31 Github Stars
DensePhrases
Open Source

DensePhrases

DensePhrases is a scalable text retrieval model designed to return phrases, sentences, passages, or documents in response to natural language queries. Built on billions of dense phrase vectors derived from the entire Wikipedia corpus, it enables real-time phrase-level answer search and supports multi-granularity retrieval for downstream tasks. The system learns dense representations of phrases to handle questions with high precision and can perform retrieval at various levels of text segmentation. Key capabilities include multi-query support, entity linking, knowledge-grounded dialogue, and slot filling. DensePhrases offers pre-trained models compatible with the Transformers library and provides smaller phrase indexes ranging from 20 to 60 GB for efficient deployment. It supports both GPU-accelerated and CPU-only inference modes. Researchers and developers can utilize the provided examples to create custom indexes, run dialogue-based document retrieval, or identify specific entities within text. The project i

ML Frameworks Search Engines
606 Github Stars