Pith Number

pith:UKSLWPQT

pith:2021:UKSLWPQT2Z3UNZ3JUCL37JYPQF

not attested not anchored not stored refs resolved

Recursively Summarizing Books with Human Feedback

Daniel M. Ziegler, Jan Leike, Jeff Wu, Long Ouyang, Nisan Stiennon, Paul Christiano, Ryan Lowe

Recursive decomposition lets models summarize entire books after humans give feedback only on short sections.

arxiv:2109.10862 v2 · 2021-09-22 · cs.CL · cs.AI · cs.LG

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{UKSLWPQT2Z3UNZ3JUCL37JYPQF}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

Our resulting model generates sensible summaries of entire books, even matching the quality of human-written summaries in a few cases (~5% of books). We achieve state-of-the-art results on the recent BookSum dataset for book-length summarization.

C2weakest assumption

That summaries of summaries retain enough information and fidelity for the final output to remain faithful to the original book when humans never see the full text.

C3one line summary

Recursive decomposition plus human feedback lets GPT-3 produce book-length summaries that reach human quality on a few cases and set new records on BookSum and NarrativeQA.

References

12 extracted · 12 resolved · 0 Pith anchors

[1] This subtask can be decomposed even further if necessary

[2] answer_directly, which returns an actual answer to the task, synthesizing the answers to subtasks In general, both decompose_if_needed and answer_directly could be learned and implemented by an ML mod 2020

[3] So gratuitously including small details is generally penalized, and omitting important details is also penalized

[4] Accuracy: All information in the summary should faithfully reﬂect the original passage

[5] We also have a fourth criteria which is primarily applicable at higher height

Formal links

3 machine-checked theorem links

Cited by

22 papers in Pith

Reinforcement Learning from Human Feedback

LLM Evaluators Recognize and Favor Their Own Generations

CALMem : Application-Layer Dual Memory for Conversational AI

Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety

RAFT: Reward rAnked FineTuning for Generative Foundation Model Alignment

Receipt and verification

First computed	2026-05-17T23:38:13.696569Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

a2a4bb3e13d67746e769a097bfa70f8148c475fb3b8ffa86c331751c38e0255a

Aliases

arxiv: 2109.10862 · arxiv_version: 2109.10862v2 · doi: 10.48550/arxiv.2109.10862 · pith_short_12: UKSLWPQT2Z3U · pith_short_16: UKSLWPQT2Z3UNZ3J · pith_short_8: UKSLWPQT

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/UKSLWPQT2Z3UNZ3JUCL37JYPQF \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: a2a4bb3e13d67746e769a097bfa70f8148c475fb3b8ffa86c331751c38e0255a

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "ecaab0f0204b68cab469b5a1be1cb0e5693ea4b755132f7c575933fc96022bdb",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.LG"
    ],
    "license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
    "primary_cat": "cs.CL",
    "submitted_at": "2021-09-22T17:34:18Z",
    "title_canon_sha256": "612ddce1253fc4e6a3ead618e152978b1fbe6ff31ad268ab30b5b82b9e65fc6e"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2109.10862",
    "kind": "arxiv",
    "version": 2
  }
}