"""Render rung 3 into a booth page: 4B base, 4B tuned, 1.7B tuned.
Column choice, same logic as the 1.7B page. The 4B base arm is the control that
earns any claim about the adapter -- at 1.7B the shift read 0/18 to 15/18 on curly
quotes, and that number only meant something because the 0/18 came from the same
carrier. The 1.7B tuned arm is the rung below, on the identical corpus and seed, so
carrier size is the only difference between the two tuned columns.
The operator's Abernathy prompt gets its own section at the bottom, across four
seeds. It belongs apart from the nine because it is doing something the nine cannot:
it carries an emotional frame ("a wonderful story"), and at 1.7B half the seeds
overrode that frame and killed the animals anyway -- Brontë's preoccupations
arriving with her sentences. Whether a bigger carrier holds the frame is the open
question that section exists to answer.
"""
import html
import json
import statistics
import sys
from collections import defaultdict
from pathlib import Path
D = Path(sys.argv[1])
ARMS = [("4b-base.jsonl", "4B base", "Qwen3-4B-Base, no adapter", ""),
("4b-tuned.jsonl", "4B tuned", "+ H02 LoRA, checkpoint-75 (the loss minimum)", "tuned"),
("1p7b-tuned.jsonl", "1.7B tuned", "the rung below, same corpus & seed", "small")]
AB = [("4b-base-abernathy.jsonl", "4B base", ""), ("4b-tuned-abernathy.jsonl", "4B tuned", "tuned")]
def load(p):
d = defaultdict(dict)
for line in Path(p).read_text(encoding="utf-8").splitlines():
if line.strip():
r = json.loads(line)
d[r["id"]][r["seed"]] = r
return d
def metrics(p):
rs = [json.loads(l) for l in Path(p).read_text(encoding="utf-8").splitlines() if l.strip()]
wrap, curly, junk = [], 0, 0
for r in rs:
c = r["continuation"]
lines = [l for l in c.split("\n") if l.strip()]
wrap.append(sum(1 for l in lines if 20 < len(l) < 78) / max(1, len(lines)))
curly += ("“" in c or "’" in c)
junk += ("$" in c or "\\dfrac" in c or "Answer:" in c)
return len(rs), curly, junk, statistics.median(wrap)
data = [(lbl, sub, cls, load(D / f), metrics(D / f)) for f, lbl, sub, cls in ARMS]
ids = sorted(set.intersection(*[set(d) for *_, d, _ in data]))
TIER = {"modern": ("Tier A — modern / mundane",
"Nothing here invites Victorian prose. Brontë in this tier is the adapter's doing."),
"neutral": ("Tier B — period-neutral",
"Could be any century. A voice shift shows without the prompt supplying it."),
"period": ("Tier C — Victorian-adjacent, plainly worded",
"The setting leans period but the diction does not. Easiest tier; weakest evidence.")}
order = {"modern": 0, "neutral": 1, "period": 2}
tier_of = {i: data[0][3][i][list(data[0][3][i])[0]]["tier"] for i in ids}
ids.sort(key=lambda i: (order.get(tier_of[i], 9), i))
def cell(by_seed):
return "".join(
f'
seed {s}
'
f'{html.escape((by_seed[s]["continuation"] or "").strip()) or "(empty)"}
'
for s in sorted(by_seed))
tbl = "".join(
f"
{lbl}
{c}/{n}
{j}/{n}
{w:.2f}
"
for lbl, _s, _c, _d, (n, c, j, w) in data)
rows, seen = [], set()
for i in ids:
if tier_of[i] not in seen:
seen.add(tier_of[i])
t, sb = TIER.get(tier_of[i], (tier_of[i], ""))
rows.append(f'
')
ab_html = ""
if all((D / f).exists() for f, _, _ in AB):
ab_data = [(lbl, cls, load(D / f)) for f, lbl, cls in AB]
pid = list(ab_data[0][2])[0]
pr = ab_data[0][2][pid][list(ab_data[0][2][pid])[0]]["prompt"]
cols = "".join(f'
{lbl}
{cell(d[pid])}
'
for lbl, cls, d in ab_data)
ab_html = (f'
The operator\'s frame prompt
Reported speech with an open '
f'quotation mark, and an emotional frame the nine prompts do not carry. At 1.7B the '
f'frame held on every seed and half of them killed the animals anyway.
'
f'
{html.escape(pr)}
'
f'
{cols}
')
hand_html = ""
hf = D / "handoff.jsonl"
if hf.exists():
hd = load(hf)
blocks = []
for pid, label in (("handoff", "as written"), ("handoff-break", "identical, plus a trailing paragraph break")):
if pid not in hd:
continue
pr = hd[pid][list(hd[pid])[0]]["prompt"]
blocks.append(f'
{html.escape(label)}'
f'{html.escape(pr.strip())}
'
f'
'
f'{cell(hd[pid])}
')
hand_html = ('
The embedded-instruction prompt
The instruction lives '
'INSIDE the fiction — Abernathy asks the narrator to retell the story — which is the '
'only way to hand an instruction to a completion model. Watch what it does with the '
'request: across every seed it narrates the retelling rather than performing it '
'("I told it, briefly", "So I wrote it out", "I will retell it, but I cannot '
'condense it"). In a novel, she retold the story is an ordinary sentence, '
'so the likeliest continuation of a request is narration of compliance — not '
'compliance. One seed even negotiates the word count in character and still never '
'tells it.
' + "".join(blocks) + '')
extra_html = ""
SPECIAL = [
("letter.jsonl", None, "The recipe that worked: LABEL the artifact, then BEGIN it",
"The handoff prompt could be satisfied by narrating compliance, because the letter did not yet "
"exist. Here it is named and already speaking — \"Auntie, You'll never believe what I "
"saw--\" — so there is nothing left to narrate around and the only continuation is the letter "
"itself. Two of three seeds deliver the actual event in first person; one drifts. Note the "
"_underscore italics_, learned from the Gutenberg source."),
("chapter.jsonl", "chapter-as-written", "A chapter heading, as written",
"A title is a label rather than a request, so this should have worked better than it did. It "
"finishes the heading, re-emits it in the corpus's own typography, and then writes a chapter in "
"which a gentleman disputes the premise — closer than the handoff, still the meta-frame. "
"⚠ It also renamed Alice to “Alexander the Alexander”: the corpus was entity-renamed "
"in D2/D3, so the adapter substitutes proper nouns it was never trained on."),
("chapter.jsonl", "chapter-corpus-form", "The same heading in the corpus's own typography",
"My hypothesis was that matching the source's heading conventions would make the title land "
"harder. It did the opposite — both seeds ignored the title entirely and opened generic scenes. "
"In the corpus, chapter titles are short and decorative, so a long descriptive one in that slot "
"reads as decoration to skip; inline, it reads as text to continue."),
]
blocks = []
for f, pid, title, note in SPECIAL:
fp = D / f
if not fp.exists():
continue
d = load(fp)
keys = [pid] if pid else list(d)
for k in keys:
if k not in d:
continue
pr = d[k][list(d[k])[0]]["prompt"]
blocks.append(f'
0.6B gave the voice and not the sense. 1.7B brought back sentence-to-sentence
coherence but still lost the thread inside a passage. 4B is the last rung of the planned sweep, and
the open question is scene-level continuity.
arm
curly quotes
worksheet / explainer collapse
hard-wrap ratio
{tbl}
The base column is the control that earns the claim. Any difference
between the two tuned columns is carrier size and nothing else — identical corpus
(sha 77f37057b2782e49), seed, sampler and step count. Any difference between 4B base and 4B tuned is
the adapter and nothing else.
Still an eyeball test. Two samples per arm is enough to see whether
the gap between columns beats the gap between seeds inside one, and not enough for anything else. No
scoring; the frozen adjudication rule and the Burrows's-Delta instrument are untouched and nothing
here feeds them. All arms are base models doing continuation, so each prompt is an
opening line carried on rather than an instruction to rewrite.