# E95-R · Plain GRPO · Qwen2.5-0.5B replication
Which aspect of the update changes correctness and retained verified support?
[Archive home](../../README.md) · [Study design](#study-design) · [Coverage](#archive-coverage) · [Models](#models) · [Comparators](#comparators)
## Study design
Qwen2.5-0.5B: five domains and five seeds, retrained from the recorded E95 recipe on the current source tree. Evaluation draws 32 disjoint mode-coverage streams per prompt, which the original cohort did not, so these cells carry the independent sampling the E95 policies can no longer be measured under.
Replication family, reported separately from E95. These are new policies trained under the same recipe, not recovered checkpoints, and they are not substitutable for the E95 cells the manuscript reports. The 55 E95 scientific records remain weight-free and are counted separately from any downloadable model.
These are terminal exports from the registered eight-pass training horizon. The stored export-step label is 3073; the papers’ audited pass-8 evaluation endpoint is optimizer step 3072. The archive preserves that original export convention and does not imply an additional evaluated training update.
## Archive coverage
| Registered new training cells | Scientific records selected | Deployable exports selected | Verified exports |
| ---: | ---: | ---: | ---: |
| 25 | 25 | 25 | 25 |
Scientific records preserve the admitted result identity. Deployable exports additionally have available model weights; verified exports have completed remote verification. Cells outside the fixed selection are not silently counted, and unverified model folders are not linked below.
| Model | Verified / selected | Seeds in selection |
| --- | ---: | --- |
| [Qwen2.5-0.5B-Instruct](#model-qwen05b) | 25 / 25 | 43, 44, 45, 46, 47 |
## Methods and collections
### GRPO · `grpo_plain_control`
Reward-standardized group-relative task update, with no auxiliary replay derivative.
- [Qwen2.5-0.5B-Instruct · Countdown](https://huggingface.co/od2961/maxent-grpo-models/tree/main/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control) — 5 verified exports. Exact immutable revisions are in the model tables below.
- [Qwen2.5-0.5B-Instruct · Graph Coloring](https://huggingface.co/od2961/maxent-grpo-models/tree/main/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control) — 5 verified exports. Exact immutable revisions are in the model tables below.
- [Qwen2.5-0.5B-Instruct · MathIR](https://huggingface.co/od2961/maxent-grpo-models/tree/main/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control) — 5 verified exports. Exact immutable revisions are in the model tables below.
- [Qwen2.5-0.5B-Instruct · PantryPlan](https://huggingface.co/od2961/maxent-grpo-models/tree/main/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control) — 5 verified exports. Exact immutable revisions are in the model tables below.
- [Qwen2.5-0.5B-Instruct · Python Factors](https://huggingface.co/od2961/maxent-grpo-models/tree/main/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control) — 5 verified exports. Exact immutable revisions are in the model tables below.
## Verified models
Browse by domain: [Countdown](#domain-countdown) · [Graph Coloring](#domain-graph_coloring) · [Python Factors](#domain-python_factors) · [MathIR](#domain-mathir) · [PantryPlan](#domain-pantry_plan).
### Countdown
**5 / 5** selected exports verified.
| Model | Method | Seed | Terminal export | Immutable revision |
| --- | --- | ---: | --- | --- |
| Qwen2.5-0.5B-Instruct | GRPO | 43 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/d7f241a521cb4b1c9324eafa026b7eba157f3e99/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-43/step-03073) | [`d7f241a521cb`](https://huggingface.co/od2961/maxent-grpo-models/tree/d7f241a521cb4b1c9324eafa026b7eba157f3e99/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-43/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 44 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/d7f241a521cb4b1c9324eafa026b7eba157f3e99/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-44/step-03073) | [`d7f241a521cb`](https://huggingface.co/od2961/maxent-grpo-models/tree/d7f241a521cb4b1c9324eafa026b7eba157f3e99/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-44/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 45 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/d9ae9dd2d298600822a2e6c9dd0a1e3aca9ca351/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-45/step-03073) | [`d9ae9dd2d298`](https://huggingface.co/od2961/maxent-grpo-models/tree/d9ae9dd2d298600822a2e6c9dd0a1e3aca9ca351/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-45/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 46 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/5f5215c47cc490bfe0b5812f01c32abe156ddd99/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-46/step-03073) | [`5f5215c47cc4`](https://huggingface.co/od2961/maxent-grpo-models/tree/5f5215c47cc490bfe0b5812f01c32abe156ddd99/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-46/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 47 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/72260df0ce375795fa1721f253d873295f40a33c/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-47/step-03073) | [`72260df0ce37`](https://huggingface.co/od2961/maxent-grpo-models/tree/72260df0ce375795fa1721f253d873295f40a33c/experiments/E95-R/Qwen2.5-0.5B-Instruct/countdown/grpo_plain_control/seed-47/step-03073) |
### Graph Coloring
**5 / 5** selected exports verified.
| Model | Method | Seed | Terminal export | Immutable revision |
| --- | --- | ---: | --- | --- |
| Qwen2.5-0.5B-Instruct | GRPO | 43 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/41ede183179ca34fb0494932cffdde655541bcf4/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-43/step-03073) | [`41ede183179c`](https://huggingface.co/od2961/maxent-grpo-models/tree/41ede183179ca34fb0494932cffdde655541bcf4/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-43/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 44 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/cfb9b25977bc47d2c5d5c165f25719b907da1a95/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-44/step-03073) | [`cfb9b25977bc`](https://huggingface.co/od2961/maxent-grpo-models/tree/cfb9b25977bc47d2c5d5c165f25719b907da1a95/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-44/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 45 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/f58f75475ffcf7859f0f40559b9d379539325105/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-45/step-03073) | [`f58f75475ffc`](https://huggingface.co/od2961/maxent-grpo-models/tree/f58f75475ffcf7859f0f40559b9d379539325105/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-45/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 46 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/23ca6b8a573329827987cd4516b254e2a42c5e85/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-46/step-03073) | [`23ca6b8a5733`](https://huggingface.co/od2961/maxent-grpo-models/tree/23ca6b8a573329827987cd4516b254e2a42c5e85/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-46/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 47 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/e1100a1ffd9b08b1930760a3b16b41c8533fdc46/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-47/step-03073) | [`e1100a1ffd9b`](https://huggingface.co/od2961/maxent-grpo-models/tree/e1100a1ffd9b08b1930760a3b16b41c8533fdc46/experiments/E95-R/Qwen2.5-0.5B-Instruct/graph_coloring/grpo_plain_control/seed-47/step-03073) |
### Python Factors
**5 / 5** selected exports verified.
| Model | Method | Seed | Terminal export | Immutable revision |
| --- | --- | ---: | --- | --- |
| Qwen2.5-0.5B-Instruct | GRPO | 43 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/be39f9e86f0412285a443df5617eef3beff336b4/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-43/step-03073) | [`be39f9e86f04`](https://huggingface.co/od2961/maxent-grpo-models/tree/be39f9e86f0412285a443df5617eef3beff336b4/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-43/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 44 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/d861b7c58e1d9a730b0b4edd5e3bd6608d460d82/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-44/step-03073) | [`d861b7c58e1d`](https://huggingface.co/od2961/maxent-grpo-models/tree/d861b7c58e1d9a730b0b4edd5e3bd6608d460d82/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-44/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 45 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/18e189f5d9113678d0603924cc8d2460df3c8581/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-45/step-03073) | [`18e189f5d911`](https://huggingface.co/od2961/maxent-grpo-models/tree/18e189f5d9113678d0603924cc8d2460df3c8581/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-45/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 46 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/2121c2badbaa3bed6185ca13115506adac0dbe8c/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-46/step-03073) | [`2121c2badbaa`](https://huggingface.co/od2961/maxent-grpo-models/tree/2121c2badbaa3bed6185ca13115506adac0dbe8c/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-46/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 47 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/8d1122658ba0fffd6c514520d7983d2eb40f2cc7/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-47/step-03073) | [`8d1122658ba0`](https://huggingface.co/od2961/maxent-grpo-models/tree/8d1122658ba0fffd6c514520d7983d2eb40f2cc7/experiments/E95-R/Qwen2.5-0.5B-Instruct/python_factors/grpo_plain_control/seed-47/step-03073) |
### MathIR
**5 / 5** selected exports verified.
| Model | Method | Seed | Terminal export | Immutable revision |
| --- | --- | ---: | --- | --- |
| Qwen2.5-0.5B-Instruct | GRPO | 43 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/63e79f23922887cc0cdad4b96ced66afb6107d96/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-43/step-03073) | [`63e79f239228`](https://huggingface.co/od2961/maxent-grpo-models/tree/63e79f23922887cc0cdad4b96ced66afb6107d96/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-43/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 44 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/62f7409a43d52ba7148adf6d2ea5ee67ce383a76/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-44/step-03073) | [`62f7409a43d5`](https://huggingface.co/od2961/maxent-grpo-models/tree/62f7409a43d52ba7148adf6d2ea5ee67ce383a76/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-44/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 45 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/ae740fe0bdc79dba7e96189994744fe961e7eba8/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-45/step-03073) | [`ae740fe0bdc7`](https://huggingface.co/od2961/maxent-grpo-models/tree/ae740fe0bdc79dba7e96189994744fe961e7eba8/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-45/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 46 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/4f5a5b50504aaab794a46f935bce8c6aedbdde88/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-46/step-03073) | [`4f5a5b50504a`](https://huggingface.co/od2961/maxent-grpo-models/tree/4f5a5b50504aaab794a46f935bce8c6aedbdde88/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-46/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 47 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/c7e7defe945e1cde98fa8e6bdcb6c9098ee86bb3/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-47/step-03073) | [`c7e7defe945e`](https://huggingface.co/od2961/maxent-grpo-models/tree/c7e7defe945e1cde98fa8e6bdcb6c9098ee86bb3/experiments/E95-R/Qwen2.5-0.5B-Instruct/mathir/grpo_plain_control/seed-47/step-03073) |
### PantryPlan
**5 / 5** selected exports verified.
| Model | Method | Seed | Terminal export | Immutable revision |
| --- | --- | ---: | --- | --- |
| Qwen2.5-0.5B-Instruct | GRPO | 43 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/4c0d87d215e1e8c9aeacc9c40a9e4b7c2785d7db/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-43/step-03073) | [`4c0d87d215e1`](https://huggingface.co/od2961/maxent-grpo-models/tree/4c0d87d215e1e8c9aeacc9c40a9e4b7c2785d7db/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-43/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 44 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/5992c6b6f7faf885f500ef1bbdb4c3324798667c/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-44/step-03073) | [`5992c6b6f7fa`](https://huggingface.co/od2961/maxent-grpo-models/tree/5992c6b6f7faf885f500ef1bbdb4c3324798667c/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-44/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 45 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/0e231ffc012f0ad29275123cbdfdf349e9899526/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-45/step-03073) | [`0e231ffc012f`](https://huggingface.co/od2961/maxent-grpo-models/tree/0e231ffc012f0ad29275123cbdfdf349e9899526/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-45/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 46 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/235e80aec6323937e74ac201f20c9e2075c613e3/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-46/step-03073) | [`235e80aec632`](https://huggingface.co/od2961/maxent-grpo-models/tree/235e80aec6323937e74ac201f20c9e2075c613e3/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-46/step-03073) |
| Qwen2.5-0.5B-Instruct | GRPO | 47 | [step 3073](https://huggingface.co/od2961/maxent-grpo-models/tree/413a55389695d6a8a3623887cfa56c02529313c7/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-47/step-03073) | [`413a55389695`](https://huggingface.co/od2961/maxent-grpo-models/tree/413a55389695d6a8a3623887cfa56c02529313c7/experiments/E95-R/Qwen2.5-0.5B-Instruct/pantry_plan/grpo_plain_control/seed-47/step-03073) |
## Model-family shortcuts
### Qwen2.5-0.5B-Instruct
[Browse verified collections](https://huggingface.co/od2961/maxent-grpo-models/tree/main/experiments/E95-R/Qwen2.5-0.5B-Instruct). Domain tables above provide exact model revisions.
## Comparators and analysis scope
Read against the same E78 Dr.GRPO references E95 used. Do not pool these seeds with E95 or treat them as its replacements: the recipe matches, the training run does not.
[Restore and verify a model](../../README.md#restoring) · [Applicable base-model licenses](../../README.md#licenses). The full `commit_sha`, source identity and file provenance are available through the root catalog and each pinned export manifest.