In an article published in the KeAi journal Plant Diversity, Dr. Lai and his team at Nanjing Forestry University in China present their new R package, phylolm.hp. This package extends the "averaged shared variance"concept to the PGLM framework. Learn more: https://t.co/ikeZIAYZKT

◼ 🔍 New study unveils PhyloLM, a groundbreaking approach that applies phylogenetic algorithms to Large Language Models (LLMs). By measuring similarities in outputs, it successfully maps relationships between 156 models & predicts their performance. A smarter way to gauge LLM

We used this analogy to create PhyloLM, a pipeline that exploits this concept to apply tools imported from genetics to LLMs : we start from a context (gene) dataset and let models sample tokens (alleles) to estimate P(allele | gene) and compute a genetic similarity matrix. 5/9

We build a distance matrix from comparing outputs of LLMs to a hundred of different contexts and build maps and trees from this distance matrix. Because PhyloLM only requires sampling very few tokens after a very short contexts the algorithm is particularly cheap to run. 1/10

Aside from plotting trees, PhyloLM similarity matrix is very versatile. For example, running a logistic regression on the distance matrix makes it possible to predict performance of new models even from unseen families with good accuracy. Here is what we got on ARC. 4/10

DHARMa v0.4.7 #rstats package for #glmm residual diagnostics now available on #CRAN. Among other things, this version includes a new residual test for #phylogenetic correlation, as well as support for the #phylolm package. Moreover, to make DHARMa more accessible for color-blind

PhyloLM : Inferring the Phylogeny of Large Language Models and Predicting their Performances in Benchmarks.

phylolm.hp: A new R package to disentangle the effects of phylogeny and environment on plant traits. No more partial R² struggles!🌱🌳🔎 #Phylogenetics #Ecology #Rstats #PlantScience @EurekAlert @KeAiPublishing

Following the long tradition of applying evolutionary concepts to cultural artefacts, @nicolas__yax developed #PhyloLM that applies genetic distance to LLMs outputs to track their relations, Not only do the method works in capturing known relaitonships https://t.co/49SVJbwplo

A PhyloLM collaborative Huggingface space is available to try the algorithm and visualize maps : https://t.co/8jXk7uFbbZ The Model Submit button has been temporarily disabled for technical reasons but you can play with the data while we fix it ! 8/10

with #phylolm by Ho & Ané.

In short, PhyloLM is a cheap and versatile algorithm that generates useful representations for LLMs that can have creative applications in pratice. 9/10 paper : https://t.co/fmkdUe6MRX colab : https://t.co/7FT4JyKKPc code : https://t.co/oOLyQRxG8E ICLR : Saturday 3pm Poster #505

法国研究团队开发了 PhyloLM 系统发育算法，通过大语言模型（LLM）的输出相似性计算系统发育距离。该方法能识别模型关联信息并预测性能，构建树枝图捕捉 111 个开源模型和 45 个封闭模型的关系，经济高效地评估 LLM 性能。

CRAN updates: HIBAG phylolm rplos usl http://t.co/y5W2NU2tPr #rstats

［秋田］もうすぐ．とりあえず今宵の紙芝居は40枚ほどできたので，ワタクシ的には「よし」としよう．計算が完了しないかもしれないが，それはそれで何とかしましょう．それにしても phylolm って使いにくいのぉ．というか，実験的な種間比較法のツールはアブナイことこの上ない．

New CRAN package phylolm with initial version 1.0 http://t.co/e88h64B8 #rstats

PhyloLM : Inferring the Phylogeny of Large Language Models and Predicting their Performances in Benchmarks https://t.co/n6CeiXew0s #devtalk

CRAN updates: dnet maptools phylolm RefManageR vbsr http://t.co/7sV9sRKPKz #rstats

@kfuku0502 l1ouのソースを見ましたが、おそらく改変すべきはphylolmまで辿る必要がありそうですね。これはなかなか大変そうです。

phylolm 1.0 – R package for Fitting Phylogenetic Linear Regression models http://t.co/8OlmXN7p