Releases · Multi-Agent-LLMs/mallm

Create update_readme.yml by @jonas-becker in #72
add workflow py script by @jonas-becker in #73
Added QA metric for GPQA, MMLU, etc. by @ItsNiklas in #74
Small Code Style Improvements by @ItsNiklas in #77
Fix readme updater by @jonas-becker in #78
feat: added batch executor for mallm by @lkaesberg in #81
ResponseGenerators: Handling prompts, extraction, and agreements by @jonas-becker in #79
More robust multichoice metric + fixed datasets by @jonas-becker in #82
Prompt improvements (+ majority consensus fix) by @jonas-becker in #85
Fix/no unlimited voting by @lkaesberg in #88
Extensive evaluation by @jonas-becker in #92
1. sort output file after finishing 2) comparable moderator by @jonas-becker in #91
feat: When there is a dataset issue, print exactly what the issue is by @jpwahle in #98
Support hf datasets by @jpwahle in #93
fix: Fix a bug where forgetting trailing slash in memory bucket leads to undesired behaviour by @jpwahle in #100
Ablation by @jonas-becker in #99
Refactor/evaluator by @lkaesberg in #90
fix: Fix a bug where the HF dataset is not sorted properly. by @jpwahle in #102
Fix/data load and debug output by @lkaesberg in #105
Feat/plotting by @lkaesberg in #106
squad metric enhancements by @jonas-becker in #107
Paraphrase types agent generator by @jonas-becker in #103
fix out full path by @jonas-becker in #109
Feat/rich output by @lkaesberg in #110
fix: unanimity is unreachable because of faulty condition by @lkaesberg in #108
Prompt changes and minor adjustments by @jonas-becker in #114
Distinct-N metric by @jonas-becker in #116
Remove dbm memory by @jonas-becker in #115
feat: added instruction templates by @lkaesberg in #118
batch executor refinements by @jonas-becker in #113
More expressive param names + flexible number of neutral agents by @jonas-becker in #119
feat: all agents generate a first draft and after that improve by @lkaesberg in #120
feat: added mmlu pro dataset by @lkaesberg in #123
feat: add musr dataset by @lkaesberg in #125
feat: add math lvl 5 dataset downloader by @lkaesberg in #126
Freely combine persona types + NoPersonaGenerator by @jonas-becker in #122
feat: added prompts for new datasets by @lkaesberg in #127
feat: add mallm command line scripts by @lkaesberg in #128
Feat/evaluator with alterations by @lkaesberg in #129
feat: added metric that checks if answer is included in response by @lkaesberg in #131
feat: added discord webhook to batch processing by @lkaesberg in #130
feat: add mmlu dataset by @lkaesberg in #132
BBQ, MoCa, MoralExceptQA Datasets + metadata field by @jonas-becker in #133
Feat/summarize decision protocol by @lkaesberg in #124
policy feedback agent by @jonas-becker in #134
Persona diversity index by @jonas-becker in #136
WinoGrande, ETHICS datasets by @jonas-becker in #137
Consensus Voting by @jonas-becker in #138
feat: add new commandline script to execute mallm batch mode by @lkaesberg in #140
feat: add challenge of final answer to test consistency by @lkaesberg in #141
update readme by @jonas-becker in #142

New Contributors

@jpwahle made their first contribution in #98

Full Changelog: v0.1.0-alpha...v1.0.0

Contributors

ItsNiklas, jpwahle, and 2 other contributors

Assets 2

17 Jun 11:54

jonas-becker

v0.1.0-alpha

7828467

v0.1.0-alpha Pre-release

Pre-release

This is the first release of MALLM (alpha).

What's Changed

Fix/setup by @lkaesberg in #1
memory redesign, more extensive output logs, refactoring, bug fixes by @jonas-becker in #2
Feat/new build system by @lkaesberg in #3
refactor: add abstract class for datasets by @lkaesberg in #4
feat: improve readme by @lkaesberg in #6
Feat/logging by @lkaesberg in #7
Feat/formatting by @lkaesberg in #8
Added GPQA Dataset, cffi dependency for linux by @ItsNiklas in #10
Tgi implementation by @jonas-becker in #9
reorganized files by @jonas-becker in #16
Multi source datasets by @jonas-becker in #19
Create unit test by @jonas-becker in #21
Fix/discussion by @lkaesberg in #17
Refactor/coordinator by @lkaesberg in #18
installable package by @jonas-becker in #25
small fixes by @jonas-becker in #26
WMT119, paraphrase types, and fixes by @jonas-becker in #28
Feat/btvote by @lkaesberg in #27
fix etpc and context by @jonas-becker in #35
fix json stringify by @jonas-becker in #36
Feat/decision protocol by @lkaesberg in #30
feat: added agent history as a chat format by @lkaesberg in #38
samples left logging and type hints by @jonas-becker in #37
stream answers to reduce memory usage by @jonas-becker in #40
Strict types by @lkaesberg in #42
Refactor/GitHub action by @lkaesberg in #45
Openai support by @jonas-becker in #47
feat: added tests for coordinator by @lkaesberg in #46
Refactor/moderator by @lkaesberg in #44
Evaluation framework by @jonas-becker in #50
Feat: fixed number of turns by @jonas-becker in #53
Improve prompts by @jonas-becker in #55
Feat baseline by @jonas-becker in #56
Added IPIP Persona Generator by @ItsNiklas in #59
fix missing hf_token, btvote not shuffled, max_samples by @jonas-becker in #57
Discussion length by @jonas-becker in #58
add pytest pre commit hook by @jonas-becker in #60
Added more decision protocols by @lkaesberg in #49
Split agree and answer by @lkaesberg in #54
refactor: introduce config dataclass to remove duplicate code and mak… by @lkaesberg in #64
Improve extraction reliability by @jonas-becker in #63
failed samples logging by @jonas-becker in #65
Stability fixes by @jonas-becker in #66

New Contributors

@lkaesberg made their first contribution in #1
@jonas-becker made their first contribution in #2

Full Changelog: https://github.com/Multi-Agent-LLMs/mallm/commits/v0.1.0-alpha

Contributors

ItsNiklas, lkaesberg, and jonas-becker

Assets 2

Releases: Multi-Agent-LLMs/mallm

v1.0.5

What's Changed

Contributors

Uh oh!

v1.0.4

What's Changed

Contributors

Uh oh!

v1.0.3

Uh oh!

v1.0.2

Uh oh!

v1.0.1

What's Changed

Contributors

Uh oh!

Public Release

What's Changed

New Contributors

Contributors

Uh oh!

v0.1.0-alpha

What's Changed

New Contributors

Contributors

Uh oh!