Update README.md

README.md

@@ -1,4 +1,87 @@
 ---
 base_model:
 - Writer/palmyra-mini
----
+---
+
+GGUF quantizations of palmyra-mini
+
+For use with llama.cpp / ollama / lm studio
+
+
+------
+
+
+# Model Card: palmyra-mini
+
+## Model Details
+
+- **Model Name:** palmyra-mini
+
+## Model Description
+
+# palmyra-mini: A Data-Driven Look at a Promising Generative Model
+
+The "palmyra-mini" model presents a compelling profile of a generative AI with distinct areas of high performance. A thorough analysis of 24 benchmarks reveals an average score of 0.3551, and more importantly, showcases a model with significant strengths in mathematical reasoning and other complex tasks for its size.
+
+## Dominance in Mathematical Reasoning
+
+Palmyra-Mini truly shines in its mathematical capabilities. The model achieved an outstanding score of **0.8180** on both **gsm8k (strict-match)** and **MATH500**. These demanding benchmarks require precise, multi-step reasoning to arrive at a correct solution, and the model's high scores are a clear indicator of its advanced logical deduction skills. This strength is further supported by a solid performance on **AMC23** with a score of **0.6000**.
+
+## Performance on Complex Benchmarks
+
+Beyond pure mathematics, the model also demonstrates competence in other complex areas. It scored **0.5259** on **BBH (get-answer)(exact_match)**, a benchmark designed to test a model's ability to handle challenging, multi-faceted problems.
+
+## Conclusion
+
+The data paints a clear picture of palmyra-minipalmyra-mini as a model with exceptional, verifiable strengths in mathematical and logical reasoning. Its high scores on some of the most challenging math benchmarks in the field are a testament to its sophisticated architecture. While there are clear opportunities for growth in other domains, the model's current capabilities establish it as a powerful and promising tool in the generative AI landscape.
+
+## Performance
+
+The model's performance is based exclusively on the data from the `pmnt-benchmarks.xlsx` spreadsheet.
+
+### Benchmark Results
+
+| benchmark                                                        |   score |
+|:-----------------------------------------------------------------|--------:|
+| gsm8k (strict-match)                                             |  0.818  |
+| minerva_math(exact_match)                                        |  0.4582 |
+| mmlu_pro(exact_match)                                            |  0.314  |
+| hendrycks_math                                                   |  0.025  |
+| ifeval (inst_level_loose_acc)                                    |  0.4688 |
+| mathqa (acc)                                                     |  0.4509 |
+| humaneval (pass@1)                                               |  0.5    |
+| BBH (get-answer)(exact_match)                                    |  0.5259 |
+| mbpp                                                             |  0.47   |
+| leadboard_musr (acc_norm)                                        |  0.3413 |
+| gpqa  lighteval gpqa diamond_pass@1:8_samples                    |  0.442  |
+| AIME24(pass@1)(avg-of-1)                                         |  0.2    |
+| AIME25(pass@1)(avg-of-1)                                         |  0.25   |
+| Livecodebench-codegen (livecodebench/code_generation_lite v4_v5) |  0.1519 |
+| AMC23                                                            |  0.6    |
+| MATH500                                                          |  0.818  |
+| Minerva                                                          |  0.2794 |
+| Olympiadbench (extractive_match)                                 |  0.3822 |
+| Codecontests (pass_rate)                                         |  0.1034 |
+| Codeforces (pass_rate)                                           |  0.3199 |
+| Taco (pass_rate)                                                 |  0.1744 |
+| APPS (all_levels)                                                |  0.0405 |
+| HMMT23 (extractive_match)                                        |  0.0333 |
+| Average                                                          |  0.3551 |
+
+### Key Strengths
+
+- **Exceptional Mathematical Reasoning:** The model demonstrates top-tier performance on mathematical benchmarks, scoring **0.8180** on both `gsm8k (strict-match)` and `MATH500`.
+- **Strong Logical Deduction:** Achieved a score of **0.6000** on `AMC23`, further highlighting its capabilities in complex problem-solving.
+- **Competence in Challenging Tasks:** Scored **0.5259** on `BBH (get-answer)(exact_match)`, indicating proficiency in handling difficult, multi-faceted problems.
+
+## Intended Use
+
+This model is well-suited for applications that require strong mathematical and logical reasoning. Potential use cases include:
+
+-   Step-by-step problem solving
+-   Educational tools for mathematics
+-   Data analysis and interpretation tasks
+
+---
+
+