Datasets:

OPENGCM
/

DeepArithmark-1

#!/usr/bin/env python3
"""
Score Hugging Face causal LMs on Deep Arithmark (and, side by side, ArithMark 3.0).
The scoring protocol follows ArithMark 3.0's (AxiomicLabs) so the two numbers
are directly comparable. This file is an independent implementation, not copied
code; it shares only imports and boilerplate with `bencharithmark-3.py`:
  * zero-shot continuation log-likelihood -- the model never generates text;
  * context and continuation tokenized separately, so the continuation cannot
    be scored through the context's tokenization;
  * `acc` sums the token log-probabilities, `acc_norm` averages them;
  * bf16, batch 32, `--max-context 1024`, by default.
On top of that it reports the profile the depth tiers are for:
  * a per-depth curve (2, 3, 4, 5 dependent operations), so a sub-1B model has
    something to climb instead of a single number pinned at chance;
  * per-template and per-topic accuracy, so a regression is attributable;
  * the share of items where `acc` and `acc_norm` disagree (a near-tie signal).
Usage
-----
    python bench_deeparithmark.py --model HuggingFaceTB/SmolLM2-135M
    python bench_deeparithmark.py --model gpt2 --compare   # ArithMark 3.0 too
    python bench_deeparithmark.py --models-file models.txt --out benchmarks/
"""

from __future__ import annotations

import argparse
import functools
import gc
import hashlib
import inspect
import json
import math
import time
from collections import defaultdict
from pathlib import Path

import torch
import torch.nn.functional as F
from huggingface_hub import hf_hub_download
from transformers import AutoModelForCausalLM, AutoTokenizer

# Some 2026 checkpoints ship remote modelling code written against transformers 4.x
# (e.g. LiquidAI/LFM3, which decorates its convolution with a hub-kernel helper).
# transformers 5 removed that helper, so the import fails before the model can load.
# The decorator only swaps in an optional fused CUDA kernel; stubbing it out leaves
# the pure-torch path, which is the same maths at the same dtype.
def _accepts_layer_idx(fn) -> bool:
    try:
        return "layer_idx" in inspect.signature(fn).parameters
    except (TypeError, ValueError):
        return True


def _install_kernel_decorator_shim() -> None:
    try:
        import transformers.integrations as ti

        if not hasattr(ti, "use_kernel_func_from_hub_with_fallback"):
            def _noop(*_args, **_kwargs):
                return lambda fn: fn

            ti.use_kernel_func_from_hub_with_fallback = _noop

        import transformers.integrations.accelerate as tia

        if not hasattr(tia, "force_accelerate_hooks"):
            def _noop_hooks(*_args, **_kwargs):
                return lambda fn: fn

            tia.force_accelerate_hooks = _noop_hooks

        import transformers.masking_utils as tmu

        if not hasattr(tmu, "create_recurrent_attention_mask"):
            # transformers 4.x used this to hand a conv/linear-attention layer the plain
            # 2-D padding mask (1 = real token). It was deleted in 5.x; the contiguity
            # logic LFM3 needs on top of it (`seq_idx`) is computed in the model itself.
            def _recurrent_mask(config=None, inputs_embeds=None, attention_mask=None,
                                past_key_values=None, position_ids=None, layer_idx=None,
                                **_kwargs):
                if attention_mask is None:
                    return None
                return attention_mask.to(torch.bool) if attention_mask.dtype != torch.bool else attention_mask

            tmu.create_recurrent_attention_mask = _recurrent_mask

        if not _accepts_layer_idx(tmu.create_causal_mask):
            # 5.5 picks the layer's mask from the hybrid cache itself, so the explicit
            # `layer_idx` LFM3 passes is redundant rather than meaningful: drop it.
            def _wrap_drop_layer_idx(fn):
                @functools.wraps(fn)
                def wrapper(*args, layer_idx=None, **kwargs):
                    del layer_idx
                    return fn(*args, **kwargs)

                return wrapper

            tmu.create_causal_mask = _wrap_drop_layer_idx(tmu.create_causal_mask)
            tmu.create_sliding_window_causal_mask = _wrap_drop_layer_idx(
                tmu.create_sliding_window_causal_mask
            )
    except Exception:  # an old transformers that has the real helper is fine too
        pass


_install_kernel_decorator_shim()

# Our own data is read from disk; this repo is not published under AxiomicLabs.
DATA_DEFAULT = "deeparithmark-dataset/data/deeparithmark-1.jsonl"
# The one AxiomicLabs reference that must stay: `--compare` downloads the upstream
# benchmark from its authors' org, which is where it actually lives. Changing this
# to another namespace would break the head-to-head.
ARITHMARK3_REPO = "AxiomicLabs/Arithmark-3.0"
ARITHMARK3_FILE = "arithmark-3.jsonl"
BATCH_SIZE = 32
MAX_CONTEXT = 1024
DEPTHS = (2, 3, 4, 5)

# every model here is under 1B parameters, which is the target of this benchmark
DEFAULT_MODELS = [
    "gpt2",
    "EleutherAI/pythia-70m",
    "HuggingFaceTB/SmolLM2-135M",
    "Qwen/Qwen2.5-0.5B",
]


def load_items(path: Path) -> list[dict]:
    items = []
    with path.open(encoding="utf-8-sig") as fh:
        for line_no, line in enumerate(fh, 1):
            if not line.strip():
                continue
            item = json.loads(line)
            label = item["label"]
            items.append(
                {
                    "id": str(item.get("ind", line_no)),
                    "ctx": item["ctx"],
                    "endings": item["endings"],
                    "label": int(label),
                    "depth": item.get("metadata", {}).get("depth"),
                    "topic": item.get("metadata", {}).get("topic"),
                    "template": item.get("metadata", {}).get("template_id"),
                }
            )
    if not items:
        raise SystemExit(f"no items in {path}")
    return items


def resolve_device(requested: str) -> str:
    if requested != "auto":
        return requested
    return "cuda" if torch.cuda.is_available() else "cpu"


def load_model(model_name: str, device: str, dtype_name: str):
    dtype = {"float32": torch.float32, "float16": torch.float16, "bfloat16": torch.bfloat16}[
        dtype_name
    ]
    if device == "cpu":
        dtype = torch.float32
    model = AutoModelForCausalLM.from_pretrained(
        model_name, dtype=dtype, trust_remote_code=True
    ).to(device)
    model.eval()
    tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True, use_fast=False)
    if tokenizer.pad_token_id is None:
        tokenizer.pad_token = tokenizer.eos_token or 0
    return model, tokenizer, dtype


@torch.inference_mode()
def score_items(model, tokenizer, device, dtype, items, batch_size: int, max_context: int):
    """Return (items, labels, raw_scores, norm_scores) -- 4 scores per item.
    Results follow the length-sorted order used for batching, so the returned
    item list is what the labels and scores line up with.
    """
    prepared = []
    for item in items:
        requests = []
        for ending in item["endings"]:
            ctx_ids = tokenizer(item["ctx"], add_special_tokens=False).input_ids
            end_ids = tokenizer(ending, add_special_tokens=False).input_ids
            if not ctx_ids or not end_ids:
                raise SystemExit(f"empty tokenization in item {item['id']}")
            ctx_ids = ctx_ids[-(max_context - len(end_ids)) :]
            requests.append((ctx_ids + end_ids, len(end_ids)))
        prepared.append((item, requests))
    # similar lengths together: less padding, fewer wasted logits
    prepared.sort(key=lambda pair: max(len(t) for t, _ in pair[1]))

    labels, raw, norm = [], [], []
    for start in range(0, len(prepared), batch_size):
        batch = prepared[start : start + batch_size]
        flats, lengths, cont_lens = [], [], []
        for _, requests in batch:
            for tokens, cont_len in requests:
                flats.append(tokens)
                lengths.append(len(tokens))
                cont_lens.append(cont_len)
        width = max(lengths)
        ids = torch.tensor(
            [t + [tokenizer.pad_token_id] * (width - len(t)) for t in flats],
            dtype=torch.long,
            device=device,
        )
        mask = torch.arange(width, device=device)[None, :] < torch.tensor(
            lengths, device=device
        )[:, None]
        starts = torch.tensor(
            [len(t) - c for t, c in zip(flats, cont_lens)], dtype=torch.long, device=device
        )
        with (
            torch.autocast("cuda", dtype=dtype) if device == "cuda" and dtype != torch.float32 else torch.no_grad()
        ):
            logits = model(ids, attention_mask=mask, use_cache=False).logits
        positions = torch.arange(1, width, device=device)[None, :]
        score_mask = (positions >= starts[:, None]) & (
            positions < torch.tensor(lengths, device=device)[:, None]
        )
        row_idx, rel = score_mask.nonzero(as_tuple=True)
        tok_pos = rel + 1
        picked = logits[row_idx, tok_pos - 1, :]
        targets = ids[row_idx, tok_pos]
        tok_ll = -F.cross_entropy(picked.float(), targets, reduction="none")
        raw_t = torch.zeros(len(flats), device=device).scatter_add_(0, row_idx, tok_ll)
        cnt_t = torch.zeros(len(flats), device=device).scatter_add_(
            0, row_idx, torch.ones_like(tok_ll)
        )
        raw_f = raw_t.cpu().tolist()
        norm_f = (raw_t / cnt_t.clamp_min(1)).cpu().tolist()
        for i, (item, _) in enumerate(batch):
            base = 4 * i
            labels.append(item["label"])
            raw.append(raw_f[base : base + 4])
            norm.append(norm_f[base : base + 4])
        del logits, ids, mask
    return [item for item, _ in prepared], labels, raw, norm


def summarize(items, labels, raw, norm) -> dict:
    def pick(scores, i):
        return max(range(4), key=lambda k: scores[i][k])

    hit_raw = [int(pick(raw, i) == labels[i]) for i in range(len(items))]
    hit_norm = [int(pick(norm, i) == labels[i]) for i in range(len(items))]

    # Soft score: the probability a proper scoring rule assigns to the correct
    # option after a softmax over the four continuations. Hard accuracy is
    # all-or-nothing (a model either lands the exact number or not), so it sits
    # at 25 % for sub-1B models; this keeps the graded signal visible.
    def soft(row, label):
        top = max(row)
        exps = [math.exp(s - top) for s in row]
        return exps[label] / sum(exps)

    soft_norm = [soft(norm[i], labels[i]) for i in range(len(items))]
    soft_by_depth = defaultdict(lambda: [0.0, 0])
    for i, item in enumerate(items):
        if item.get("depth"):
            soft_by_depth[item["depth"]][0] += soft_norm[i]
            soft_by_depth[item["depth"]][1] += 1

    def pct(correct: int, n: int) -> float:
        return 100.0 * correct / n if n else 0.0

    by_depth = defaultdict(lambda: [0, 0])
    by_topic = defaultdict(lambda: [0, 0])
    by_template = defaultdict(lambda: [0, 0])
    for i, item in enumerate(items):
        if item.get("depth"):
            by_depth[item["depth"]][0] += hit_norm[i]
            by_depth[item["depth"]][1] += 1
        by_topic[item.get("topic") or "n/a"][0] += hit_norm[i]
        by_topic[item.get("topic") or "n/a"][1] += 1
        by_template[item.get("template") or "n/a"][0] += hit_norm[i]
        by_template[item.get("template") or "n/a"][1] += 1

    n = len(items)
    return {
        "n": n,
        "acc": pct(sum(hit_raw), n),
        "acc_norm": pct(sum(hit_norm), n),
        "soft": 100.0 * sum(soft_norm) / n,
        "disagreements": sum(1 for a, b in zip(hit_raw, hit_norm) if a != b),
        "by_depth": {
            d: {
                "acc_norm": pct(v[0], v[1]),
                "soft": 100.0 * soft_by_depth[d][0] / soft_by_depth[d][1],
                "n": v[1],
            }
            for d, v in sorted(by_depth.items())
        },
        "by_topic": {
            t: {"acc_norm": pct(v[0], v[1]), "n": v[1]} for t, v in sorted(by_topic.items())
        },
        "by_template": {
            t: {"acc_norm": pct(v[0], v[1]), "n": v[1]} for t, v in sorted(by_template.items())
        },
    }


def print_report(name, params, res) -> None:
    print(f"\n{name}  ({params:,} parameters)")
    print(f"  Deep Arithmark score (acc_norm)  {res['acc_norm']:6.2f}%")
    print(f"  soft score (P[correct])          {res['soft']:6.2f}%")
    print(f"  raw token-sum accuracy (acc)     {res['acc']:6.2f}%")
    if res["by_depth"]:
        print("  depth curve (acc_norm / soft):")
        for depth, vals in res["by_depth"].items():
            print(
                f"    depth {depth}: {vals['acc_norm']:6.2f}% / {vals['soft']:6.2f}%"
                f"  (n={vals['n']})"
            )
    print("  by template:")
    for template, vals in sorted(res["by_template"].items(), key=lambda kv: -kv[1]["acc_norm"]):
        print(f"    {template:34s} {vals['acc_norm']:6.2f}%  (n={vals['n']})")


def arithmark3_path() -> Path:
    return Path(hf_hub_download(ARITHMARK3_REPO, ARITHMARK3_FILE, repo_type="dataset"))


def run_model(model_name, args, device, deep_items, am3_items):
    model, tokenizer, dtype = load_model(model_name, device, args.dtype)
    params = sum(p.numel() for p in model.parameters())
    out = {"model": model_name, "params": params, "protocol": {
        "dtype": str(dtype), "batch_size": args.batch_size,
        "max_context": args.max_context, "shots": 0,
    }}
    t0 = time.perf_counter()
    ordered, labels, raw, norm = score_items(
        model, tokenizer, device, dtype, deep_items, args.batch_size, args.max_context
    )
    res = summarize(ordered, labels, raw, norm)
    out["deeparithmark"] = res
    print_report(model_name, params, res)
    if args.dump_predictions:
        target = Path(args.dump_predictions)
        target.parent.mkdir(parents=True, exist_ok=True)
        with target.open("a", encoding="utf-8") as fh:
            for i, item in enumerate(ordered):
                fh.write(
                    json.dumps(
                        {
                            "model": model_name,
                            "id": item["id"],
                            "depth": item.get("depth"),
                            "template": item.get("template"),
                            "label": labels[i],
                            "raw_pick": max(range(4), key=lambda k: raw[i][k]),
                            "norm_pick": max(range(4), key=lambda k: norm[i][k]),
                            "raw": raw[i],
                            "norm": norm[i],
                        }
                    )
                    + "\n"
                )
    if am3_items is not None:
        ordered, labels, raw, norm = score_items(
            model, tokenizer, device, dtype, am3_items, args.batch_size, args.max_context
        )
        am3 = summarize(ordered, labels, raw, norm)
        out["arithmark3"] = am3
        print(f"  ArithMark 3.0 (acc_norm)         {am3['acc_norm']:6.2f}%   [same weights, same protocol]")
        print(f"  drop                             {am3['acc_norm'] - res['acc_norm']:+6.2f} points")
    out["seconds"] = round(time.perf_counter() - t0, 2)
    del model, tokenizer
    gc.collect()
    if device == "cuda":
        torch.cuda.empty_cache()
    return out


def main() -> None:
    ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
    ap.add_argument("--model", action="append", help="HF model id; repeatable")
    ap.add_argument("--data", default=DATA_DEFAULT)
    ap.add_argument("--compare", action="store_true", help="also score ArithMark 3.0")
    ap.add_argument("--arithmark3", help="local path to arithmark-3.jsonl")
    ap.add_argument("--batch-size", type=int, default=BATCH_SIZE)
    ap.add_argument("--max-context", type=int, default=MAX_CONTEXT)
    ap.add_argument("--dtype", choices=("float32", "float16", "bfloat16"), default="bfloat16")
    ap.add_argument("--device", choices=("auto", "cpu", "cuda"), default="auto")
    ap.add_argument("--out", default="benchmarks")
    ap.add_argument("--no-save", action="store_true")
    ap.add_argument(
        "--dump-predictions",
        help="write per-item scores to this JSONL path for error analysis",
    )
    args = ap.parse_args()

    torch.set_float32_matmul_precision("high")
    if torch.cuda.is_available():
        torch.backends.cuda.matmul.allow_tf32 = True

    data_path = Path(args.data)
    if not data_path.exists():
        raise SystemExit(f"missing {data_path}; run `python deeparithmark.py build`")
    deep_items = load_items(data_path)
    sha = hashlib.sha256(data_path.read_bytes()).hexdigest()
    device = resolve_device(args.device)
    print(f"device: {device} | items: {len(deep_items)} | sha256: {sha[:16]}...")

    am3_items = None
    if args.compare or args.arithmark3:
        am3_file = Path(args.arithmark3) if args.arithmark3 else arithmark3_path()
        am3_items = load_items(am3_file)
        print(f"comparing against ArithMark 3.0 ({len(am3_items)} items): {am3_file}")

    models = args.model or DEFAULT_MODELS
    results = []
    for name in models:
        try:
            results.append(run_model(name, args, device, deep_items, am3_items))
        except Exception as exc:  # a broken hub repo should not kill the sweep
            print(f"\n{name}: FAILED -- {str(exc).splitlines()[0]}")

    if results:
        print("\n" + "=" * 84)
        print(f"{'model':40s} {'deep':>8s} {'soft':>8s} {'arith3':>8s}")
        for r in results:
            am3 = r.get("arithmark3", {}).get("acc_norm")
            print(
                f"{r['model']:40s} {r['deeparithmark']['acc_norm']:7.2f}%"
                f" {r['deeparithmark']['soft']:7.2f}%"
                + (f" {am3:7.2f}%" if am3 is not None else "       -")
            )
        print("=" * 84)

    if not args.no_save and results:
        out_dir = Path(args.out)
        out_dir.mkdir(parents=True, exist_ok=True)
        payload = {
            "dataset": {"name": "deeparithmark-1", "path": str(data_path), "sha256": sha,
                        "items": len(deep_items)},
            "protocol": {
                "shots": 0, "dtype": args.dtype, "batch_size": args.batch_size,
                "max_context": args.max_context, "metric": "acc_norm",
            },
            "results": results,
        }
        target = out_dir / f"deeparithmark_{len(results)}models.json"
        target.write_text(json.dumps(payload, indent=2), encoding="utf-8")
        print(f"\nresults -> {target}")


if __name__ == "__main__":
    main()

	#!/usr/bin/env python3
	"""
	Score Hugging Face causal LMs on Deep Arithmark (and, side by side, ArithMark 3.0).

	The scoring protocol follows ArithMark 3.0's (AxiomicLabs) so the two numbers
	are directly comparable. This file is an independent implementation, not copied
	code; it shares only imports and boilerplate with `bencharithmark-3.py`:

	* zero-shot continuation log-likelihood -- the model never generates text;
	* context and continuation tokenized separately, so the continuation cannot
	be scored through the context's tokenization;
	* `acc` sums the token log-probabilities, `acc_norm` averages them;
	* bf16, batch 32, `--max-context 1024`, by default.

	On top of that it reports the profile the depth tiers are for:

	* a per-depth curve (2, 3, 4, 5 dependent operations), so a sub-1B model has
	something to climb instead of a single number pinned at chance;
	* per-template and per-topic accuracy, so a regression is attributable;
	* the share of items where `acc` and `acc_norm` disagree (a near-tie signal).

	Usage
	-----
	python bench_deeparithmark.py --model HuggingFaceTB/SmolLM2-135M
	python bench_deeparithmark.py --model gpt2 --compare # ArithMark 3.0 too
	python bench_deeparithmark.py --models-file models.txt --out benchmarks/
	"""

	from __future__ import annotations

	import argparse
	import functools
	import gc
	import hashlib
	import inspect
	import json
	import math
	import time
	from collections import defaultdict
	from pathlib import Path

	import torch
	import torch.nn.functional as F
	from huggingface_hub import hf_hub_download
	from transformers import AutoModelForCausalLM, AutoTokenizer

	# Some 2026 checkpoints ship remote modelling code written against transformers 4.x
	# (e.g. LiquidAI/LFM3, which decorates its convolution with a hub-kernel helper).
	# transformers 5 removed that helper, so the import fails before the model can load.
	# The decorator only swaps in an optional fused CUDA kernel; stubbing it out leaves
	# the pure-torch path, which is the same maths at the same dtype.
	def _accepts_layer_idx(fn) -> bool:
	try:
	return "layer_idx" in inspect.signature(fn).parameters
	except (TypeError, ValueError):
	return True


	def _install_kernel_decorator_shim() -> None:
	try:
	import transformers.integrations as ti

	if not hasattr(ti, "use_kernel_func_from_hub_with_fallback"):
	def _noop(_args, *_kwargs):
	return lambda fn: fn

	ti.use_kernel_func_from_hub_with_fallback = _noop

	import transformers.integrations.accelerate as tia

	if not hasattr(tia, "force_accelerate_hooks"):
	def _noop_hooks(_args, *_kwargs):
	return lambda fn: fn

	tia.force_accelerate_hooks = _noop_hooks

	import transformers.masking_utils as tmu

	if not hasattr(tmu, "create_recurrent_attention_mask"):
	# transformers 4.x used this to hand a conv/linear-attention layer the plain
	# 2-D padding mask (1 = real token). It was deleted in 5.x; the contiguity
	# logic LFM3 needs on top of it (`seq_idx`) is computed in the model itself.
	def _recurrent_mask(config=None, inputs_embeds=None, attention_mask=None,
	past_key_values=None, position_ids=None, layer_idx=None,
	**_kwargs):
	if attention_mask is None:
	return None
	return attention_mask.to(torch.bool) if attention_mask.dtype != torch.bool else attention_mask

	tmu.create_recurrent_attention_mask = _recurrent_mask

	if not _accepts_layer_idx(tmu.create_causal_mask):
	# 5.5 picks the layer's mask from the hybrid cache itself, so the explicit
	# `layer_idx` LFM3 passes is redundant rather than meaningful: drop it.
	def _wrap_drop_layer_idx(fn):
	@functools.wraps(fn)
	def wrapper(args, layer_idx=None, *kwargs):
	del layer_idx
	return fn(args, *kwargs)

	return wrapper

	tmu.create_causal_mask = _wrap_drop_layer_idx(tmu.create_causal_mask)
	tmu.create_sliding_window_causal_mask = _wrap_drop_layer_idx(
	tmu.create_sliding_window_causal_mask
	)
	except Exception: # an old transformers that has the real helper is fine too
	pass


	_install_kernel_decorator_shim()

	# Our own data is read from disk; this repo is not published under AxiomicLabs.
	DATA_DEFAULT = "deeparithmark-dataset/data/deeparithmark-1.jsonl"
	# The one AxiomicLabs reference that must stay: `--compare` downloads the upstream
	# benchmark from its authors' org, which is where it actually lives. Changing this
	# to another namespace would break the head-to-head.
	ARITHMARK3_REPO = "AxiomicLabs/Arithmark-3.0"
	ARITHMARK3_FILE = "arithmark-3.jsonl"
	BATCH_SIZE = 32
	MAX_CONTEXT = 1024
	DEPTHS = (2, 3, 4, 5)

	# every model here is under 1B parameters, which is the target of this benchmark
	DEFAULT_MODELS = [
	"gpt2",
	"EleutherAI/pythia-70m",
	"HuggingFaceTB/SmolLM2-135M",
	"Qwen/Qwen2.5-0.5B",
	]


	def load_items(path: Path) -> list[dict]:
	items = []
	with path.open(encoding="utf-8-sig") as fh:
	for line_no, line in enumerate(fh, 1):
	if not line.strip():
	continue
	item = json.loads(line)
	label = item["label"]
	items.append(
	{
	"id": str(item.get("ind", line_no)),
	"ctx": item["ctx"],
	"endings": item["endings"],
	"label": int(label),
	"depth": item.get("metadata", {}).get("depth"),
	"topic": item.get("metadata", {}).get("topic"),
	"template": item.get("metadata", {}).get("template_id"),
	}
	)
	if not items:
	raise SystemExit(f"no items in {path}")
	return items


	def resolve_device(requested: str) -> str:
	if requested != "auto":
	return requested
	return "cuda" if torch.cuda.is_available() else "cpu"


	def load_model(model_name: str, device: str, dtype_name: str):
	dtype = {"float32": torch.float32, "float16": torch.float16, "bfloat16": torch.bfloat16}[
	dtype_name
	]
	if device == "cpu":
	dtype = torch.float32
	model = AutoModelForCausalLM.from_pretrained(
	model_name, dtype=dtype, trust_remote_code=True
	).to(device)
	model.eval()
	tokenizer = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True, use_fast=False)
	if tokenizer.pad_token_id is None:
	tokenizer.pad_token = tokenizer.eos_token or 0
	return model, tokenizer, dtype


	@torch.inference_mode()
	def score_items(model, tokenizer, device, dtype, items, batch_size: int, max_context: int):
	"""Return (items, labels, raw_scores, norm_scores) -- 4 scores per item.

	Results follow the length-sorted order used for batching, so the returned
	item list is what the labels and scores line up with.
	"""
	prepared = []
	for item in items:
	requests = []
	for ending in item["endings"]:
	ctx_ids = tokenizer(item["ctx"], add_special_tokens=False).input_ids
	end_ids = tokenizer(ending, add_special_tokens=False).input_ids
	if not ctx_ids or not end_ids:
	raise SystemExit(f"empty tokenization in item {item['id']}")
	ctx_ids = ctx_ids[-(max_context - len(end_ids)) :]
	requests.append((ctx_ids + end_ids, len(end_ids)))
	prepared.append((item, requests))
	# similar lengths together: less padding, fewer wasted logits
	prepared.sort(key=lambda pair: max(len(t) for t, _ in pair[1]))

	labels, raw, norm = [], [], []
	for start in range(0, len(prepared), batch_size):
	batch = prepared[start : start + batch_size]
	flats, lengths, cont_lens = [], [], []
	for _, requests in batch:
	for tokens, cont_len in requests:
	flats.append(tokens)
	lengths.append(len(tokens))
	cont_lens.append(cont_len)
	width = max(lengths)
	ids = torch.tensor(
	[t + [tokenizer.pad_token_id] * (width - len(t)) for t in flats],
	dtype=torch.long,
	device=device,
	)
	mask = torch.arange(width, device=device)[None, :] < torch.tensor(
	lengths, device=device
	)[:, None]
	starts = torch.tensor(
	[len(t) - c for t, c in zip(flats, cont_lens)], dtype=torch.long, device=device
	)
	with (
	torch.autocast("cuda", dtype=dtype) if device == "cuda" and dtype != torch.float32 else torch.no_grad()
	):
	logits = model(ids, attention_mask=mask, use_cache=False).logits
	positions = torch.arange(1, width, device=device)[None, :]
	score_mask = (positions >= starts[:, None]) & (
	positions < torch.tensor(lengths, device=device)[:, None]
	)
	row_idx, rel = score_mask.nonzero(as_tuple=True)
	tok_pos = rel + 1
	picked = logits[row_idx, tok_pos - 1, :]
	targets = ids[row_idx, tok_pos]
	tok_ll = -F.cross_entropy(picked.float(), targets, reduction="none")
	raw_t = torch.zeros(len(flats), device=device).scatter_add_(0, row_idx, tok_ll)
	cnt_t = torch.zeros(len(flats), device=device).scatter_add_(
	0, row_idx, torch.ones_like(tok_ll)
	)
	raw_f = raw_t.cpu().tolist()
	norm_f = (raw_t / cnt_t.clamp_min(1)).cpu().tolist()
	for i, (item, _) in enumerate(batch):
	base = 4 * i
	labels.append(item["label"])
	raw.append(raw_f[base : base + 4])
	norm.append(norm_f[base : base + 4])
	del logits, ids, mask
	return [item for item, _ in prepared], labels, raw, norm


	def summarize(items, labels, raw, norm) -> dict:
	def pick(scores, i):
	return max(range(4), key=lambda k: scores[i][k])

	hit_raw = [int(pick(raw, i) == labels[i]) for i in range(len(items))]
	hit_norm = [int(pick(norm, i) == labels[i]) for i in range(len(items))]

	# Soft score: the probability a proper scoring rule assigns to the correct
	# option after a softmax over the four continuations. Hard accuracy is
	# all-or-nothing (a model either lands the exact number or not), so it sits
	# at 25 % for sub-1B models; this keeps the graded signal visible.
	def soft(row, label):
	top = max(row)
	exps = [math.exp(s - top) for s in row]
	return exps[label] / sum(exps)

	soft_norm = [soft(norm[i], labels[i]) for i in range(len(items))]
	soft_by_depth = defaultdict(lambda: [0.0, 0])
	for i, item in enumerate(items):
	if item.get("depth"):
	soft_by_depth[item["depth"]][0] += soft_norm[i]
	soft_by_depth[item["depth"]][1] += 1

	def pct(correct: int, n: int) -> float:
	return 100.0 * correct / n if n else 0.0

	by_depth = defaultdict(lambda: [0, 0])
	by_topic = defaultdict(lambda: [0, 0])
	by_template = defaultdict(lambda: [0, 0])
	for i, item in enumerate(items):
	if item.get("depth"):
	by_depth[item["depth"]][0] += hit_norm[i]
	by_depth[item["depth"]][1] += 1
	by_topic[item.get("topic") or "n/a"][0] += hit_norm[i]
	by_topic[item.get("topic") or "n/a"][1] += 1
	by_template[item.get("template") or "n/a"][0] += hit_norm[i]
	by_template[item.get("template") or "n/a"][1] += 1

	n = len(items)
	return {
	"n": n,
	"acc": pct(sum(hit_raw), n),
	"acc_norm": pct(sum(hit_norm), n),
	"soft": 100.0 * sum(soft_norm) / n,
	"disagreements": sum(1 for a, b in zip(hit_raw, hit_norm) if a != b),
	"by_depth": {
	d: {
	"acc_norm": pct(v[0], v[1]),
	"soft": 100.0 * soft_by_depth[d][0] / soft_by_depth[d][1],
	"n": v[1],
	}
	for d, v in sorted(by_depth.items())
	},
	"by_topic": {
	t: {"acc_norm": pct(v[0], v[1]), "n": v[1]} for t, v in sorted(by_topic.items())
	},
	"by_template": {
	t: {"acc_norm": pct(v[0], v[1]), "n": v[1]} for t, v in sorted(by_template.items())
	},
	}


	def print_report(name, params, res) -> None:
	print(f"\n{name} ({params:,} parameters)")
	print(f" Deep Arithmark score (acc_norm) {res['acc_norm']:6.2f}%")
	print(f" soft score (P[correct]) {res['soft']:6.2f}%")
	print(f" raw token-sum accuracy (acc) {res['acc']:6.2f}%")
	if res["by_depth"]:
	print(" depth curve (acc_norm / soft):")
	for depth, vals in res["by_depth"].items():
	print(
	f" depth {depth}: {vals['acc_norm']:6.2f}% / {vals['soft']:6.2f}%"
	f" (n={vals['n']})"
	)
	print(" by template:")
	for template, vals in sorted(res["by_template"].items(), key=lambda kv: -kv[1]["acc_norm"]):
	print(f" {template:34s} {vals['acc_norm']:6.2f}% (n={vals['n']})")


	def arithmark3_path() -> Path:
	return Path(hf_hub_download(ARITHMARK3_REPO, ARITHMARK3_FILE, repo_type="dataset"))


	def run_model(model_name, args, device, deep_items, am3_items):
	model, tokenizer, dtype = load_model(model_name, device, args.dtype)
	params = sum(p.numel() for p in model.parameters())
	out = {"model": model_name, "params": params, "protocol": {
	"dtype": str(dtype), "batch_size": args.batch_size,
	"max_context": args.max_context, "shots": 0,
	}}
	t0 = time.perf_counter()
	ordered, labels, raw, norm = score_items(
	model, tokenizer, device, dtype, deep_items, args.batch_size, args.max_context
	)
	res = summarize(ordered, labels, raw, norm)
	out["deeparithmark"] = res
	print_report(model_name, params, res)
	if args.dump_predictions:
	target = Path(args.dump_predictions)
	target.parent.mkdir(parents=True, exist_ok=True)
	with target.open("a", encoding="utf-8") as fh:
	for i, item in enumerate(ordered):
	fh.write(
	json.dumps(
	{
	"model": model_name,
	"id": item["id"],
	"depth": item.get("depth"),
	"template": item.get("template"),
	"label": labels[i],
	"raw_pick": max(range(4), key=lambda k: raw[i][k]),
	"norm_pick": max(range(4), key=lambda k: norm[i][k]),
	"raw": raw[i],
	"norm": norm[i],
	}
	)
	+ "\n"
	)
	if am3_items is not None:
	ordered, labels, raw, norm = score_items(
	model, tokenizer, device, dtype, am3_items, args.batch_size, args.max_context
	)
	am3 = summarize(ordered, labels, raw, norm)
	out["arithmark3"] = am3
	print(f" ArithMark 3.0 (acc_norm) {am3['acc_norm']:6.2f}% [same weights, same protocol]")
	print(f" drop {am3['acc_norm'] - res['acc_norm']:+6.2f} points")
	out["seconds"] = round(time.perf_counter() - t0, 2)
	del model, tokenizer
	gc.collect()
	if device == "cuda":
	torch.cuda.empty_cache()
	return out


	def main() -> None:
	ap = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
	ap.add_argument("--model", action="append", help="HF model id; repeatable")
	ap.add_argument("--data", default=DATA_DEFAULT)
	ap.add_argument("--compare", action="store_true", help="also score ArithMark 3.0")
	ap.add_argument("--arithmark3", help="local path to arithmark-3.jsonl")
	ap.add_argument("--batch-size", type=int, default=BATCH_SIZE)
	ap.add_argument("--max-context", type=int, default=MAX_CONTEXT)
	ap.add_argument("--dtype", choices=("float32", "float16", "bfloat16"), default="bfloat16")
	ap.add_argument("--device", choices=("auto", "cpu", "cuda"), default="auto")
	ap.add_argument("--out", default="benchmarks")
	ap.add_argument("--no-save", action="store_true")
	ap.add_argument(
	"--dump-predictions",
	help="write per-item scores to this JSONL path for error analysis",
	)
	args = ap.parse_args()

	torch.set_float32_matmul_precision("high")
	if torch.cuda.is_available():
	torch.backends.cuda.matmul.allow_tf32 = True

	data_path = Path(args.data)
	if not data_path.exists():
	raise SystemExit(f"missing {data_path}; run `python deeparithmark.py build`")
	deep_items = load_items(data_path)
	sha = hashlib.sha256(data_path.read_bytes()).hexdigest()
	device = resolve_device(args.device)
	print(f"device: {device} \| items: {len(deep_items)} \| sha256: {sha[:16]}...")

	am3_items = None
	if args.compare or args.arithmark3:
	am3_file = Path(args.arithmark3) if args.arithmark3 else arithmark3_path()
	am3_items = load_items(am3_file)
	print(f"comparing against ArithMark 3.0 ({len(am3_items)} items): {am3_file}")

	models = args.model or DEFAULT_MODELS
	results = []
	for name in models:
	try:
	results.append(run_model(name, args, device, deep_items, am3_items))
	except Exception as exc: # a broken hub repo should not kill the sweep
	print(f"\n{name}: FAILED -- {str(exc).splitlines()[0]}")

	if results:
	print("\n" + "=" * 84)
	print(f"{'model':40s} {'deep':>8s} {'soft':>8s} {'arith3':>8s}")
	for r in results:
	am3 = r.get("arithmark3", {}).get("acc_norm")
	print(
	f"{r['model']:40s} {r['deeparithmark']['acc_norm']:7.2f}%"
	f" {r['deeparithmark']['soft']:7.2f}%"
	+ (f" {am3:7.2f}%" if am3 is not None else " -")
	)
	print("=" * 84)

	if not args.no_save and results:
	out_dir = Path(args.out)
	out_dir.mkdir(parents=True, exist_ok=True)
	payload = {
	"dataset": {"name": "deeparithmark-1", "path": str(data_path), "sha256": sha,
	"items": len(deep_items)},
	"protocol": {
	"shots": 0, "dtype": args.dtype, "batch_size": args.batch_size,
	"max_context": args.max_context, "metric": "acc_norm",
	},
	"results": results,
	}
	target = out_dir / f"deeparithmark_{len(results)}models.json"
	target.write_text(json.dumps(payload, indent=2), encoding="utf-8")
	print(f"\nresults -> {target}")


	if __name__ == "__main__":
	main()