JesseMarkowitzandClaude Opus 5
e5617b86ba
The pipeline's embed-and-cluster step is dead, and this commit holds both the evidence for that and the step proposed to replace it. Predicaments. Scenes are re-described as "what the person is up against", with no names, jobs or places, then embedded and clustered (redescribe.py, topic_words.py, topic_share.py). The pilot chose qwen3:14b over 3b by reading both side by side. Two defects the pilot exposed are fixed: split.py missed titles in quotes and a contents subtitle after a dash, so three stories had been merged into their neighbours, and strip_names.py read New York place names as people. The corrected corpus is probe/v2 (97 stories, 839 scenes); carry_summaries.py reuses the 829 unchanged v1 summaries. Topic share fell from 20% to 13% at k=60, short of the pre-registered 10%. Hand references. Three corpora were read scene by scene and written up by hand, under the same prompt rules the local models get, as a baseline to judge them against: O. Henry (probe/v2/claude, 839 scenes, 20 situations), Wharton's Descent of Man (probe/wharton, 262 scenes, 16 groups) and Jacobs's The Lady of the Barge (probe/jacobs, 157 scenes, 19 groups). Each has its own README and a readable page. No inference was used for any of them. Catalogue. probe/catalogue maps every hand group in the three references onto 36 situation entries, with an answer key per corpus and one recurrence rule applied to all three. classify.py assigns a scene one entry or none, leave-one-corpus- out; score.py checks it against the key, with a self-test on random labels. Why clustering is out: hand-written predicaments, embedded and clustered exactly as the model's were, agree with the hand grouping at ARI 0.05 — no better than the 14B text's 0.07. Better rewriting cannot rescue it. Embeddings cannot even shortlist: the hand label is the nearest entry 13% of the time and in the top 8 half the time. The classification runs are not here. The dev and test runs are pre-registered in probe/catalogue/README.md with the bar set beforehand, and are blocked on the inference host, whose GPU has fallen off the PCIe bus three times. The 30-scene partial output in out/ is not a result. Review page. The situation review is now a browser page rather than JSON edited by hand (review_page.py, review_page_logic.cjs with Node tests, format schema v2). It has never been rendered in a real browser. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014BygvsUXV9eU6oHkTCkKZ1
16 lines
3.4 KiB
JSON
16 lines
3.4 KiB
JSON
[
|
|
{"chunk":67,"story":5,"summary":"Late at night one co-owner of a house and a business sits hating his partner, who has discovered his embezzlement and ordered him to leave both by morning.","predicament":"A man found out in wrongdoing is ordered by his partner to give up everything and leave by morning."},
|
|
{"chunk":68,"story":5,"summary":"The honest partner refuses to cover the theft quietly, offers the exile two hundred pounds, and mentions the servants are away until morning.","predicament":"A man offered a harsh but fair way out of his disgrace wants a quieter bargain the other will not give."},
|
|
{"chunk":69,"story":5,"summary":"The exile demands the matter be hushed up, draws a sword from the wall, and stabs his partner when he refuses.","predicament":"A man facing ruin threatens the only person who can expose him, and must choose whether to go through with it."},
|
|
{"chunk":70,"story":5,"summary":"Sickened, the killer turns out the lights, flees to the gate, and sees a constable coming up the road.","predicament":"A man who has just done something terrible in rage panics and must steady himself when a figure of the law appears."},
|
|
{"chunk":71,"story":5,"summary":"He bids the constable good night, returns inside, and gathers his cash and valuables, reckoning he has eight hours' start.","predicament":"A man with a crime behind him and a few hours' start must decide what to take and where to run."},
|
|
{"chunk":72,"story":5,"summary":"He plans his escape, weighing a crowded district against the seaside, and finds every choice may mean life or death.","predicament":"A fugitive planning his escape finds that every choice might be the one that gets him caught."},
|
|
{"chunk":73,"story":5,"summary":"At one o'clock he hears creaks on the stairs, and fears the dead man is moving about the house.","predicament":"A man alone with what he has done hears noises in the dark and fears it has come back for him."},
|
|
{"chunk":74,"story":5,"summary":"Finding the library door open, he locks someone inside, then realises it is a burglar he can blame for the killing.","predicament":"A guilty man realises a stranger's blunder gives him someone else to blame."},
|
|
{"chunk":75,"story":5,"summary":"He fires through the door, raises the alarm, and the police break in and seize the terrified intruder.","predicament":"A man framing an innocent person must act the outraged victim in front of the law."},
|
|
{"chunk":76,"story":5,"summary":"The intruder swears the body was there when he arrived; the sergeant uncovers it, and the killer feigns horror.","predicament":"A man must play the shocked innocent while the person he framed pleads the truth."},
|
|
{"chunk":77,"story":5,"summary":"The killer names his friend and accuses the intruder, who accuses him back; the sergeant seems to side with the killer.","predicament":"Someone falsely accused points to the real culprit, who must hold his nerve and keep up his story."},
|
|
{"chunk":78,"story":5,"summary":"The sergeant takes the pistol, handcuffs the killer without warning, and sends for a doctor, since the victim is not dead.","predicament":"A man who thought he had got away with it learns the one witness who could hang him is still alive."},
|
|
{"chunk":79,"story":5,"summary":"The dying partner opens his eyes and speaks; the sergeant takes notes and leads the killer away.","predicament":"A man whose deceit has collapsed can only watch as the truth comes out and he is taken away."}
|
|
]
|