Add gold deliverable workbooks (tasks 7, 9); rewrite README; update task table

.gitattributes

@@ -301,3 +301,5 @@ reference_folder/filesystem/5.[[:space:]]Analysis/Merger[[:space:]]Model/!Archiv
 reference_folder/filesystem/5.[[:space:]]Analysis/Merger[[:space:]]Model/Project[[:space:]]Helix[[:space:]]-[[:space:]]Merger[[:space:]]Model[[:space:]]-[[:space:]]vF.xlsx filter=lfs diff=lfs merge=lfs -text
 reference_folder/filesystem/6.[[:space:]]Presentations/!Archive/Project[[:space:]]Helix[[:space:]]-[[:space:]]Acquisition[[:space:]]Proposal[[:space:]]v1.pptx filter=lfs diff=lfs merge=lfs -text
 reference_folder/filesystem/6.[[:space:]]Presentations/Project[[:space:]]Helix[[:space:]]-[[:space:]]Acquisition[[:space:]]Proposal[[:space:]]v2.pptx filter=lfs diff=lfs merge=lfs -text
+deliverable_files/task_helix_dhr_ilmn_07/Project[[:space:]]Helix[[:space:]]-[[:space:]]DCF[[:space:]]Model[[:space:]]-[[:space:]]Task[[:space:]]7[[:space:]]Gold[[:space:]]Standard[[:space:]](ex-china).xlsx filter=lfs diff=lfs merge=lfs -text
+deliverable_files/task_helix_dhr_ilmn_09/Project[[:space:]]Helix[[:space:]]-[[:space:]]Merger[[:space:]]Model[[:space:]]-[[:space:]]Task[[:space:]]9[[:space:]]Gold[[:space:]]Standard.xlsx filter=lfs diff=lfs merge=lfs -text

README.md

@@ -6,14 +6,14 @@ configs:
         path: data/train-*
 ---
 
-# finance-data — DHR / ILMN M&A Tasks (GDPval-style)
+# finance-data — DHR / ILMN M&A Analysis Tasks
 
-Real-world investment-banking analysis tasks set in a (fictional) strategic M&A
-project where Danaher (DHR) explores acquiring Illumina (ILMN). Each task gives
-an agent a prompt plus a full deal-room of reference materials (financial
-models, spreadsheets, PDFs, research) and grades the answer against a
-per-criterion rubric. Format follows [openai/gdpval](https://huggingface.co/datasets/openai/gdpval),
-adapted where our data is richer.
+Investment-banking analysis tasks set in a (fictional) strategic M&A project
+where Danaher (DHR) explores acquiring Illumina (ILMN). Each task gives an
+agent a prompt plus a full deal-room of reference materials (financial models,
+spreadsheets, PDFs, research) and grades the answer against a per-criterion
+rubric. Tasks that require modifying a workbook also ship the expert's gold
+workbook as a deliverable file.
 
 ## Layout
 
@@ -22,11 +22,11 @@ data/                 # the task table (parquet, 3 rows)
 reference_folder/     # the shared deal-room every task works from
 ├── filesystem/       #   models, filings, research, deal docs (xlsx/pdf/pptx/…)
 └── .apps_data/       #   app state (mail, calendar, chat) for the simulated environment
+deliverable_files/    # gold deliverable workbooks, one subfolder per task
 ```
 
-Unlike GDPval's per-task `reference_files`, all tasks here share **one
-reference folder** (they are set in the same deal). The `reference_folder`
-column holds its repo-relative path.
+All tasks share **one reference folder** — they are set in the same deal. The
+`reference_folder` column holds its repo-relative path.
 
 ## Columns
 
@@ -37,11 +37,10 @@ column holds its repo-relative path.
 | `occupation` | string | `Investment Banking Analyst` |
 | `prompt` | string | Self-contained task instructions (includes scenario framing) |
 | `reference_folder` | string | Repo-relative path to the shared deal-room (`reference_folder`) |
-| `deliverable_files` | list[string] | Repo-relative paths to gold deliverable files (currently empty — see note) |
+| `deliverable_files` | list[string] | Repo-relative paths to the gold deliverable workbook(s); empty for answer-only tasks |
 | `gold_response` | string | The expert gold answer (text) |
-| `gold_file_names` | list[string] | Names of gold deliverable files referenced by the task author |
 | `rubric_pretty` | string | Human-readable rubric |
-| `rubric_json` | string | JSON-encoded rubric (native format, see below) |
+| `rubric_json` | string | JSON-encoded rubric (see below) |
 
 ## Rubric format (`rubric_json`)
 
@@ -57,16 +56,10 @@ A JSON array; each entry is one grading criterion with fields:
 - `dependencies` — criteria that must pass for this one to be meaningful
 - `criterion_type`, `evaluation_scope`, `expected_file_type`, `human_rating`, `tags`, `universal`
 
-This is richer than GDPval's flat score/criterion checklist: criteria carry
-numeric tolerances, weights, primary-objective flags, and inter-criterion
-dependencies.
+## Intended harness
 
-## Notes
-
-- Task `task_helix_dhr_ilmn_09` expects a modified Excel workbook as its
-  deliverable; its gold workbook (`Project Helix - Merger Model - Task 9 Gold
-  Standard.xlsx`, listed in `gold_file_names`) is not included as we only hold
-  the world's starting state. `deliverable_files` is empty for all tasks.
-- Intended harness: populate a sandboxed environment from `reference_folder`,
-  run the agent with file/spreadsheet/PDF tools, grade the final answer with an
-  LLM judge applying each rubric criterion (pass iff all primary criteria pass).
+Populate a sandboxed environment from `reference_folder`, run the agent with
+file/spreadsheet/PDF tools, and grade the final answer with an LLM judge that
+applies each rubric criterion within its stated tolerance. A task passes iff
+every primary-objective criterion passes. For tasks with `deliverable_files`,
+the gold workbook shows the expected end-state of the modified model.

data/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:7a3d0fde5b4a69732cf46e0c84943a556ef2ef50ee9c9b322eda01131f35a8e8
-size 48636
+oid sha256:ad7ec62e539256fb60cfaefcaace5188742f6077a9e71f4465532dc750e8c0c7
+size 48568

deliverable_files/task_helix_dhr_ilmn_07/Project Helix - DCF Model - Task 7 Gold Standard (ex-china).xlsx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3c6c7c407729066c0f3ecf1381caa7549b6f1fbbe1e78e10f0397fa82355daa
+size 141483

deliverable_files/task_helix_dhr_ilmn_09/Project Helix - Merger Model - Task 9 Gold Standard.xlsx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed1b4697e66aed2a307b8ac674630b39ed663a408a175b1f2c7f8b46cff29c14
+size 146643