v0.4.0

Release with a pip package

What's Changed

• Added BLEU metric by @ArtemVazh in #213
• New model loading scripts by @rvashurin in #211
• Claim-level evaluation for Chinese and Arabic languages by @alfekka and @ruixing76
• Instruct-tuned models support and reflexive methods by @rvashurin in #221
• Blackbox models support via OpenAI API by @yobeen in #228
• Rejection rate limit for PRR calculation by @ArtemVazh in #230

Full Changelog: https://github.com/IINemo/lm-polygraph/compare/v0.3.0..v0.4.0

v.0.3.0

Release with a pip package

What's Changed

• updated configs for gsm8k by @ArtemVazh in #194
• Fix entropy calculation by @rvashurin in #197
• Fix gsm8k config by @ArtemVazh in #196
• Claim level by @cant-access-rediska0123 in #192
• Fix claims extraction by @cant-access-rediska0123 in #203
• OpenAI chat try-except added, prompt fixed by @Aktsvigun in #204
• Fix warning by @rvashurin in #205
• Automatic publish to PyPI when creating new release by @SpeedOfMagic in #209
• Monotonic calibration by @rvashurin in #199

Full Changelog: v0.2.2...v0.3.0