Pith Number

pith:JLQU7SVE

pith:2026:JLQU7SVEBQXVMED4TOJ2L4AIZT

not attested not anchored not stored refs resolved

Revisiting DAgger in the Era of LLM-Agents

Bo Dai, Changhao Li, Chao Zhang, Chenxiao Gao, Jiawei Huang, Niao He, Rushi Qiang

DAgger with turn-level interpolation mitigates covariate shift in multi-turn LLM agents while retaining dense teacher supervision.

arxiv:2605.12913 v1 · 2026-05-13 · 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{JLQU7SVEBQXVMED4TOJ2L4AIZT}

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

DAgger-style training with turn-level interpolation of student and teacher policies mitigates covariate shift while retaining dense teacher supervision, producing +3.9 and +3.6 point gains on SWE-bench Verified for 4B and 8B models respectively.

C2weakest assumption

That a reliable teacher policy remains available and affordable to query at every training step and that the environment state can be reset or continued after mixed student-teacher actions without introducing new distribution shifts.

C3one line summary

DAgger-style training with turn-level policy interpolation raises 4B and 8B LLM agents to 27.3% and 29.8% on SWE-bench Verified, beating several larger published systems.

References

61 extracted · 61 resolved · 22 Pith anchors

[1] GPT-4 Technical Report 2023 · arXiv:2303.08774

[2] On-policy distillation of language models: Learning from self-generated mistakes 2024

[3] Dream: Deep research evaluation with agentic metrics.arXiv preprint arXiv:2602.18940, 2026 2026

[4] Swe-rebench: An automated pipeline for task collection and decontaminated evaluation of software engineering agents 2025

[5] Qwen3-Coder-Next Technical Report 2026 · arXiv:2603.00729

Receipt and verification

First computed	2026-05-18T03:09:10.428168Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

4ae14fcaa40c2f56107c9b93a5f008ccfecfc096119f3a1b84becc00c187ca59

Aliases

arxiv: 2605.12913 · arxiv_version: 2605.12913v1 · doi: 10.48550/arxiv.2605.12913 · pith_short_12: JLQU7SVEBQXV · pith_short_16: JLQU7SVEBQXVMED4 · pith_short_8: JLQU7SVE

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/JLQU7SVEBQXVMED4TOJ2L4AIZT \
  | 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: 4ae14fcaa40c2f56107c9b93a5f008ccfecfc096119f3a1b84becc00c187ca59

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "7444eb4d392079d6da7cbd69c325f4d369540ce2a50f21a9d27125570f0d9c2d",
    "cross_cats_sorted": [],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-05-13T02:40:28Z",
    "title_canon_sha256": "766daa8d1047e84f86de4b371a86636845d7f49480e0028602a03dcd011c9521"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.12913",
    "kind": "arxiv",
    "version": 1
  }
}