pith. sign in
Pith Number

pith:HVONQ4JD

pith:2026:HVONQ4JDUBXOGVDVSBGHTGUV7A
not attested not anchored not stored refs pending

Maximizing Mutual Information Between Prompt and Response Improves LLM Performance With No Additional Data

Haoran Li, Hyunji Nam, Natasha Jaques

Maximizing mutual information between prompts and responses lets LLMs improve personalization and problem solving without new data or oversight.

arxiv:2603.19294 v3 · 2026-03-10 · cs.LG · cs.AI · cs.CL

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{HVONQ4JDUBXOGVDVSBGHTGUV7A}

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

using Direct Preference Optimization (DPO) to learn from this paired data maximizes pointwise conditional mutual information (MI), under the base LLM, between prompts and model responses

C2weakest assumption

That responses generated by conditioning on random unrelated prompts form sufficiently informative negatives such that DPO on the resulting pairs actually maximizes the desired pointwise conditional mutual information and produces measurable downstream gains

C3one line summary

MIPO constructs contrastive preference pairs from correct versus random prompts and uses DPO to maximize mutual information between prompts and responses, producing 3-40% gains on personalization and 1-18% on math tasks without new data or oversight.

Formal links

2 machine-checked theorem links

Receipt and verification
First computed 2026-05-29T01:04:37.182899Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

3d5cd87123a06ee35475904c799a95f81df9e4268d472b01ee1ea9baf990069a

Aliases

arxiv: 2603.19294 · arxiv_version: 2603.19294v3 · doi: 10.48550/arxiv.2603.19294 · pith_short_12: HVONQ4JDUBXO · pith_short_16: HVONQ4JDUBXOGVDV · pith_short_8: HVONQ4JD
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/HVONQ4JDUBXOGVDVSBGHTGUV7A \
  | 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: 3d5cd87123a06ee35475904c799a95f81df9e4268d472b01ee1ea9baf990069a
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "5baac0306ed355e8ac347a9fed882e498f70cbff830c853bd7aa626a90944280",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CL"
    ],
    "license": "http://creativecommons.org/licenses/by-nc-nd/4.0/",
    "primary_cat": "cs.LG",
    "submitted_at": "2026-03-10T21:00:05Z",
    "title_canon_sha256": "8fdfba85b3e55e2ca0764ff3f5f1036d3511394e5b141dabdb9c42fffa7dc91c"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2603.19294",
    "kind": "arxiv",
    "version": 3
  }
}