Spaces:
Running
Running
Upload pages/index.md with huggingface_hub
Browse files- pages/index.md +5 -2
pages/index.md
CHANGED
|
@@ -25,5 +25,8 @@ Paper: [OpenReview gLRqqQOM9K](https://openreview.net/forum?id=gLRqqQOM9K)
|
|
| 25 |
Claims 1–2 now have independent exact checks plus a source-pinned audit of the
|
| 26 |
authors' own upward-propagation code. Claim 5 has a matched reduced-scale real
|
| 27 |
LM run: CARS accepted 2/10 XML documents, ARS 1/10, and RS 0/10 under the same
|
| 28 |
-
checkpoint, prompt, seed, token cap, and call budget.
|
| 29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
| 25 |
Claims 1–2 now have independent exact checks plus a source-pinned audit of the
|
| 26 |
authors' own upward-propagation code. Claim 5 has a matched reduced-scale real
|
| 27 |
LM run: CARS accepted 2/10 XML documents, ARS 1/10, and RS 0/10 under the same
|
| 28 |
+
checkpoint, prompt, seed, token cap, and call budget. Claim 6 is additionally
|
| 29 |
+
verified from all 174 released Llama-8B BV4/SLIA traces. Claim 3 has a
|
| 30 |
+
source-pinned released-trace falsification of the reported Acrylates magnitude,
|
| 31 |
+
with two same-pipeline controls. Claim 4 remains queued because the released
|
| 32 |
+
PDDL trial inventory is noncanonical.
|