Comments (1)
Hi @sofi444, thanks for your interest in our work. This should technically be possible if the retrieval is good enough to find articles which are highly relevant to the underlying fact.
Our package currently uses a two stage retrieval process, the first is article selection using SQL (ref) followed by BM25 ranking of the passages (ref). You could potentially combine it into one (very expensive with BM25Okapi
), or use fuzzier matching in stage 1 to find relevant articles.
Note that fact checking for harder facts is an iterative process, and an ideal system probably needs to hop over multiple articles to reliably fact-check (our package is quite far from supporting that).
Hope this helps, closing the issue for now but feel free to re-open if you have more questions! :)
from factscore.
Related Issues (15)
- Improving logging, show different levels
- Better ranking between models HOT 1
- MAX_LENGTH is not defined in retrieval HOT 2
- License?
- Donβt understand how true_score and false_score are represented in logits HOT 1
- Move abstain detection outside the atomic fact generator
- Using unlabeled data to generate atomic facts and retrieving evidence HOT 1
- FActScore can be slow to run
- Having issues on using LLAMA-7B model HOT 2
- estimated cost for API before running API HOT 1
- Add citation / bibtex to README HOT 1
- Create HF weight diffs for LLAMA model and add it to the README
- improve processing of Perplexity AI outputs HOT 2
- Return respond rate and number of facts together with FActScore HOT 1
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
π Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. πππ
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google β€οΈ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from factscore.