Simplify fixed proposals and remove cached combination scores
Browse files- README.md +4 -2
- manifest.json +2 -2
README.md
CHANGED
|
@@ -10,7 +10,7 @@ tags:
|
|
| 10 |
|
| 11 |
These are the two core inference models for *The Alien Space of Science: Sampling Coherent but Cognitively Unavailable Research Directions*. Run them with the [public code](https://github.com/alejandrohdez00/alien-space-of-science).
|
| 12 |
|
| 13 |
-
Training data, paper artifacts, and
|
| 14 |
|
| 15 |
## Contents
|
| 16 |
|
|
@@ -32,8 +32,10 @@ CPU and CUDA are supported. Sampling does not use an LLM API. Reconstruction is
|
|
| 32 |
|
| 33 |
Coherence is negative mean atom-token NLL over all set permutations, excluding EOS. Availability is the median of the top ten author cosine similarities. Fusion is `(1-beta)*z(coherence) + beta*z(-availability)` over the candidate population. Each checkpoint has its own token mapping; preserve the files together.
|
| 34 |
|
| 35 |
-
## Training
|
| 36 |
|
| 37 |
Coherence uses 136,827 training and 15,203 validation records from 15,203 papers. Availability uses 43,077,647 usable training and 4,786,404 validation subset/complement records; the training bank contains 39,791 authors from a corpus with 46,329 author atom profiles. Availability validation is by example, not by held-out author.
|
| 38 |
|
|
|
|
|
|
|
| 39 |
Source: the authors' paper corpus and model training runs. File sizes and checksums are in `manifest.json`. The model weights and repository code are released under the MIT license; see `LICENSE`.
|
|
|
|
| 10 |
|
| 11 |
These are the two core inference models for *The Alien Space of Science: Sampling Coherent but Cognitively Unavailable Research Directions*. Run them with the [public code](https://github.com/alejandrohdez00/alien-space-of-science).
|
| 12 |
|
| 13 |
+
Training data, paper artifacts, and fixed proposals are in [alien-data](https://huggingface.co/datasets/alejandrohdez/alien-data).
|
| 14 |
|
| 15 |
## Contents
|
| 16 |
|
|
|
|
| 32 |
|
| 33 |
Coherence is negative mean atom-token NLL over all set permutations, excluding EOS. Availability is the median of the top ten author cosine similarities. Fusion is `(1-beta)*z(coherence) + beta*z(-availability)` over the candidate population. Each checkpoint has its own token mapping; preserve the files together.
|
| 34 |
|
| 35 |
+
## Training and limitations
|
| 36 |
|
| 37 |
Coherence uses 136,827 training and 15,203 validation records from 15,203 papers. Availability uses 43,077,647 usable training and 4,786,404 validation subset/complement records; the training bank contains 39,791 authors from a corpus with 46,329 author atom profiles. Availability validation is by example, not by held-out author.
|
| 38 |
|
| 39 |
+
The models recombine the fixed 273-atom LLM-research catalogue. Availability measures corpus-relative proximity to author repertoires; it is not proof of originality, feasibility, or future scientific value. Decoder outputs are proposals to assess. CPU and GPU calculations can differ numerically.
|
| 40 |
+
|
| 41 |
Source: the authors' paper corpus and model training runs. File sizes and checksums are in `manifest.json`. The model weights and repository code are released under the MIT license; see `LICENSE`.
|
manifest.json
CHANGED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"files": {
|
| 3 |
"README.md": {
|
| 4 |
-
"bytes":
|
| 5 |
-
"sha256": "
|
| 6 |
},
|
| 7 |
"availability_model/COMPLETE": {
|
| 8 |
"bytes": 44,
|
|
|
|
| 1 |
{
|
| 2 |
"files": {
|
| 3 |
"README.md": {
|
| 4 |
+
"bytes": 2881,
|
| 5 |
+
"sha256": "91854db2858ce45d0866bd6648e2bd17200e851c2fb7d79d55c24b0ea99505fe"
|
| 6 |
},
|
| 7 |
"availability_model/COMPLETE": {
|
| 8 |
"bytes": 44,
|