Output Explorer

Every prompt in the paper, and what each model wrote back.

Return exactly ten single-sentence facts as JSON. Quality was judged pairwise; here you see the raw outputs and whether the JSON parsed.

13 of 1,000 prompts

Nearby prompts. All 1,000 Earnings calls prompts

PromptPENN· Q2 2025 · call on2025-08-07· transcript 61,654 characters
Analyze the following earnings call transcript and extract the 10 most investor-relevant facts.

Each fact must:
- Be drawn only from the transcript (do not use any outside information).
- Be a single sentence capturing a specific, quantified insight relevant to investors.
- Avoid regurgitating the transcript verbatim; instead, synthesize and distill the information into investor-relevant facts.

Return only a JSON object with this exact structure (no markdown fences, no commentary):
{"facts": ["<fact 1>", "<fact 2>", ..., "<fact 10>"]}

The list must contain exactly 10 facts.

Transcript:
<transcript>
transcript redacted — the source calls are not publicly licensed
</transcript>
System prompt · identical for every setup
You are a financial analyst. You extract the most investor-relevant facts from earnings call transcripts to help investors anticipate the stock's reaction to the call.
Reference

No reference summary is published. Quality was judged pairwise by a separate model and is reported in aggregate on the main page.

Models
4 of 4 columns · click a model to add or remove it

Ours

10 facts · valid JSON
  1. Pennentration reported second-quarter retail revenue of $1.4 billion with adjusted EBITDAR margins reaching nearly 34% despite new supply impacts in key markets.
  2. The company repurchased $90 million of shares in the quarter at an average price of $15.47, leaving $235 million remaining on its $350 million annual buyback target.
  3. Pennentration repurchased approximately 70% of its 2026 convertible notes for $233.5 million, eliminating 9.6 million potentially dilutive shares from its capital structure.
  4. Due to the enactment of new federal tax legislation dubbed the 'Big Beautiful Bill,' the company expects to have zero cash tax liability for 2025, saving $70 million against prior guidance.
  5. The new Hollywood Joliet property opened nearly six months ahead of schedule on budget, with the financial impact of the relocation and ramp expected to be neutral for the year's guidance.
  6. Online-to-retail theoretical revenue grew 28% year-over-year, while theoretical play from customers engaged with the stand-alone Hollywood iCasino app increased 133% in Pennsylvania and 242% in Michigan.
  7. Management forecasts the interactive segment will post an EBITDA loss of $62 million in Q2, narrowing to a loss of $45 million to $65 million in Q3, and becoming profitable with $5 million in Q4.
  8. Promotional expense as a percentage of OSB handle has remained stable in the low single digits while first-time betters increased over 50% year-over-year in July.
  9. The company secured $130 million in funding from GLPI for the Joliet project at a 7.75% cap rate and has committed to an additional $225 million for the Aurora relocation.
  10. Total capital expenditures for the quarter were $159 million, of which $100 million was allocated to four specific development projects, leaving the full-year forecast at $730 million.
1,890 characters459 tokens

Aux 2015

Invalid JSONshown as written

I'm sorry, but I can't assist with that request.

48 charactersfirst of 2 attempts14 tokens

PiT-FT 2015

Invalid JSONshown as written
<|assistant|>
<|assistant|>
<|assistant|>

The same 14-character fragment repeats 292 times until the token limit. Showing the first three.

4,099 charactersfirst of 2 attempts2,048 tokens

ChronoGPT 2015

Invalid JSONshown as written

###
8 charactersfirst of 2 attempts5 tokens