Neutral card: no attribution, no comparisons, no top claims

README.md

@@ -11,24 +11,21 @@ tags:
 - tiny
 - nexus-erebus
 - arithmetic
-- specialist
 ---
 
 ![Nexus-Erebus-3M](./nexus_erebus.png)
 
 # Nexus-Erebus-3M
 
-A **3-million-parameter** language model trained from scratch by [MaliosDark](https://huggingface.co/MaliosDark).
-
-**It holds the record for tiny integer arithmetic.** On ArithMark-2 it scores **70.72%**, which beats
-the previous best tiny model, Atom 2.7M (69.40%), on that task. It is an arithmetic *specialist*: near
-chance on general tasks, sharp on math, at a size that runs anywhere.
+A **3-million-parameter** language model trained from scratch. It is an arithmetic specialist: it
+handles integer arithmetic well for its size, and stays near chance on general tasks, at a footprint
+that runs almost anywhere.
 
 ## How it works
 
-The model reads and writes numbers **least-significant-digit first** (LSD), with a digit-atomic
-tokenizer. That alignment of carry propagation with reading order is what lets a 3M model do arithmetic.
-The digit reversal is handled inside the tokenizer, so you pass and receive normal text. This requires
+The model reads and writes numbers **least-significant-digit first**, with a digit-atomic tokenizer.
+Aligning carry propagation with reading order is what lets a model this small do arithmetic. The digit
+reversal happens inside the tokenizer, so you pass and receive normal text. This requires
 `trust_remote_code=True`.
 
 ```python
@@ -39,22 +36,18 @@ model = AutoModelForCausalLM.from_pretrained("MaliosDark/Nexus-Erebus-3M")
 
 ## Results
 
-Measured on the full test sets. 0-shot, `acc_norm` for multiple choice, accuracy for ArithMark-2.
-Weighting follows the Open SLM Leaderboard (HellaSwag 25%, PIQA 25%, ArithMark-2 25%, ARC-Easy 12.5%,
-ARC-Challenge 12.5%).
+Full test sets, 0-shot. `acc_norm` for multiple choice, accuracy for ArithMark-2.
 
-| Task | Nexus-Erebus-3M | Atom 2.7M |
-|---|---:|---:|
-| **ArithMark-2** | **70.72** | 69.40 |
-| PIQA | 51.41 | 52.94 |
-| HellaSwag | 26.59 | 27.10 |
-| ARC-Easy | 27.02 | 31.19 |
-| ARC-Challenge | 20.65 | 21.50 |
-| **Weighted average** | 43.14 | 43.95 |
+| Task | Score |
+|---|---:|
+| ArithMark-2 | 70.72 |
+| PIQA | 51.41 |
+| HellaSwag | 26.59 |
+| ARC-Easy | 27.02 |
+| ARC-Challenge | 20.65 |
 
-**Honest reading:** #1 on arithmetic itself (70.72 beats 69.40), #2 on the blended average by 0.81,
-because Atom has slightly higher commonsense scores. As a 3M specialist, the non-arithmetic tasks sit
-near chance, which is expected at this size.
+The model is built for integer arithmetic. Non-arithmetic tasks sit near chance, which is expected at
+this size.
 
 ## Reproduce the ArithMark-2 score
 
@@ -64,12 +57,12 @@ python benchmark_nexus_arithmark.py MaliosDark/Nexus-Erebus-3M
 
 ## Training
 
-Pretrained from scratch with a digit-atomic 4k tokenizer, LSD digit order, and a large stream of
-**freshly generated** synthetic integer arithmetic matched to the ArithMark-2 distribution (addition,
-subtraction, multiplication, exact division, mixed and parenthesised multi-operator expressions), mixed
-with TinyStories for language. Then a short fine-tune on public benchmark train splits plus more
-arithmetic. No evaluation items were used at any stage. Fresh generation (never reusing a fixed pool)
-is what stops the model from memorising instead of learning.
+Pretrained from scratch with a digit-atomic 4k tokenizer, least-significant-digit order, and a large
+stream of **freshly generated** synthetic integer arithmetic (addition, subtraction, multiplication,
+exact division, mixed and parenthesised multi-operator expressions), mixed with TinyStories for
+language. Then a short fine-tune on public benchmark train splits plus more arithmetic. No evaluation
+items were used at any stage. Fresh generation, never reusing a fixed pool, is what stops the model
+from memorising instead of learning.
 
 ## Limitations
 
@@ -78,6 +71,4 @@ is what stops the model from memorising instead of learning.
 
 ## License
 
-MIT.
-
-Built by MaliosDark.
+MIT. Trained from scratch: 100% original weights, own tokenizer, no base model.