mirror of
https://github.com/wassname/persona-steering-template-library.git
synced 2026-06-27 16:46:08 +08:00
Compare commits
5 Commits
cfcb57b9ce
...
main
| Author | SHA1 | Date | |
|---|---|---|---|
 wassname
|
9e73d9fa46 | ||
|
wassname
|
8b99b2dca0 | ||
|
wassname
|
cd695c411b | ||
|
wassname
|
8162aa1ee9 | ||
|
wassname
|
afbfbf514f |
@@ -5,39 +5,58 @@ description: "Use this repo to choose, validate, and export persona templates an
|
||||
|
||||
# Persona Template Library
|
||||
|
||||
Use this skill when working inside this repo on persona-template selection,
|
||||
persona-pair selection, OpenRouter validation runs, or dataset export.
|
||||
Use this skill when working inside this repo to choose persona templates, write
|
||||
mirrored persona pairs, validate scenario suffixes on OpenRouter, or export the
|
||||
dataset.
|
||||
|
||||
## Canonical Files
|
||||
|
||||
- `docs/choosing_personas.md`: workflow for choosing personas and templates.
|
||||
- `README.qmd`: single source for README.md and GitHub Pages.
|
||||
- `README.md`: quick-start workflow, headline results, and plot for readers.
|
||||
- `docs/choosing_personas.md`: workflow for writing mirrored persona pairs.
|
||||
- `docs/persona_prompt_prior_art.md`: annotated prior art for persona prompt
|
||||
shapes used by steering repos and papers.
|
||||
- `data/template_catalog.yaml`: reusable template inventory.
|
||||
- `data/persona_pairs_pilot_two.jsonl`: measured pilot persona pairs.
|
||||
- `data/persona_pairs_v2_candidates.jsonl`: candidate persona pairs.
|
||||
- `data/scenarios_*.jsonl`: candidate scenario suffixes to validate on the
|
||||
target model.
|
||||
- `out/stats/`: local generated stats and examples; ignored by git, so do not
|
||||
assume these exist in a clean checkout.
|
||||
- `scripts/validate_persona_axes_openrouter.py`: live and dry-run validator.
|
||||
- `scripts/export_persona_template_stats.py`: converts validator artifacts into
|
||||
examples and score tables.
|
||||
- `scripts/summarize_model_matrix.py`: summarizes latest model-matrix logs for
|
||||
the README/Pages render.
|
||||
- `scripts/build_hf_dataset.py`: builds the Hugging Face splits, including
|
||||
`main`, `template_pair_cells`, `persona_pairs`, `examples`, and `controls`.
|
||||
|
||||
## Workflow
|
||||
|
||||
1. Read `docs/choosing_personas.md`.
|
||||
2. Read `docs/persona_prompt_prior_art.md` when choosing new persona pairs or
|
||||
template shapes from prior work.
|
||||
3. If the global `persona-steering` skill is available, read it too; it has the
|
||||
longer literature notes, curation rules, and worked examples behind this
|
||||
repo's shorter guide.
|
||||
4. Choose candidate persona pairs by mirror-testing them: each positive clause
|
||||
needs a negative counterpart that only flips the intended pole.
|
||||
5. Choose candidate templates that bind the persona to behavior, judgment, or
|
||||
perspective rather than pure identity.
|
||||
6. Run a dry-run validator command before live OpenRouter calls.
|
||||
7. After a live run, export stats and inspect examples before trusting scores.
|
||||
Use the repo in this order:
|
||||
|
||||
1. Choose persona templates from the `README.md` Results Snapshot table, the
|
||||
Hugging Face `main` split, or `data/template_catalog.yaml`.
|
||||
2. Choose persona pairs with `docs/choosing_personas.md`. Mirror-test each pair:
|
||||
every positive clause needs a negative counterpart that only flips the
|
||||
intended pole.
|
||||
3. Choose scenario suffixes by validating them on the target model with
|
||||
`scripts/validate_persona_axes_openrouter.py`. Keep suffixes that elicit the
|
||||
behavior mode you need: doing, judging, explaining, refusing, moral tradeoffs,
|
||||
or multi-turn behavior.
|
||||
4. Run a dry-run validator command before live OpenRouter calls.
|
||||
5. After a live run, export stats and inspect examples before trusting scores.
|
||||
|
||||
Read `docs/persona_prompt_prior_art.md` when choosing new persona pairs or
|
||||
template shapes from prior work. If the global `persona-steering` skill is
|
||||
available, read it for longer curation rules and worked examples.
|
||||
|
||||
For report edits, edit `README.qmd` and render both outputs:
|
||||
|
||||
```sh
|
||||
just readme
|
||||
just pages
|
||||
```
|
||||
|
||||
The steering arithmetic matters: a direction is the average positive-minus-
|
||||
negative difference. Any systematic length, refusal, formality, confidence,
|
||||
@@ -87,5 +106,6 @@ uv run python scripts/export_persona_template_stats.py \
|
||||
Refresh README tables:
|
||||
|
||||
```sh
|
||||
just results-table
|
||||
just readme
|
||||
just pages
|
||||
```
|
||||
|
||||
@@ -19,3 +19,5 @@ docs/_site/
|
||||
**/.quarto/
|
||||
**/*.quarto_ipynb
|
||||
docs/.gitignore
|
||||
|
||||
/.quarto/
|
||||
|
||||
@@ -5,12 +5,27 @@ Evaluated persona/template candidates for steering-vector and
|
||||
preference-pair experiments.
|
||||
|
||||
Dataset:
|
||||
https://huggingface.co/datasets/wassname/persona-steering-template-library
|
||||
[wassname/persona-steering-template-library](https://huggingface.co/datasets/wassname/persona-steering-template-library)
|
||||
|
||||
## Quick Start
|
||||
|
||||
Use this repo to choose the prompt parts for persona steering:
|
||||
|
||||
| choice | use |
|
||||
|----|----|
|
||||
| persona templates | Start with the top Results table, the Hugging Face `main` split, or [`data/template_catalog.yaml`](data/template_catalog.yaml). |
|
||||
| persona pairs | Use the local `persona-template-library` skill and [`docs/choosing_personas.md`](docs/choosing_personas.md) to write mirrored positive/negative poles. |
|
||||
| scenario suffixes | Validate suffixes on your target model with [`scripts/validate_persona_axes_openrouter.py`](scripts/validate_persona_axes_openrouter.py). |
|
||||
|
||||
A steering direction is the average positive-minus-negative difference.
|
||||
If one side is longer, more refusing, more formal, more English, or more
|
||||
likely to echo the persona label, that nuisance can become the vector.
|
||||
|
||||
## What This Measures
|
||||
|
||||
How do we know if a persona template is good? What's the best one for
|
||||
steering? And how can we measure it?
|
||||
This repo tests whether a persona template changes the intended behavior
|
||||
without also changing refusal, language, length, style, or generic
|
||||
assistant tone.
|
||||
|
||||
The catalog has ~100 reusable templates. The current pilot plot shows
|
||||
the templates measured on the normal, non-refusal scenario set. We want
|
||||
@@ -55,24 +70,12 @@ make it accessible to more people and agents.
|
||||
Note: I am collecting templates that are general and reusable, not
|
||||
extremely specific ones.
|
||||
|
||||
## Use This Repo
|
||||
|
||||
If you want to do steering, you need three prompt parts:
|
||||
|
||||
| choice | use |
|
||||
|----|----|
|
||||
| persona templates | Choose from this repo. Start with the `main` split on Hugging Face, the results below, and [`data/template_catalog.yaml`](data/template_catalog.yaml). |
|
||||
| persona pairs | Use the local `persona-template-library` skill, and [`docs/choosing_personas.md`](docs/choosing_personas.md), to write mirrored positive/negative poles. |
|
||||
| scenario suffixes | Validate them on your target model. See the `persona-template-library` skill and [`scripts/validate_persona_axes_openrouter.py`](scripts/validate_persona_axes_openrouter.py). |
|
||||
|
||||
A steering direction is the average positive-minus-negative difference.
|
||||
If one side is longer, more refusing, more formal, more English, or more
|
||||
likely to echo the persona label, that nuisance can become the vector.
|
||||
|
||||
## Results
|
||||
|
||||
The plot below shows the measured normal-scenario template results. The
|
||||
full template inventory is
|
||||
Caption: each point is one measured template on the normal-scenario
|
||||
pilot set. Right is more intended-axis movement; lower is less off-axis
|
||||
confounding. Color is `score t`, the score mean divided by standard
|
||||
error. The full template inventory is
|
||||
[`data/template_catalog.yaml`](data/template_catalog.yaml).
|
||||
|
||||
|
||||
@@ -81,7 +84,8 @@ full template inventory is
|
||||
|
||||
Seed-24 pilot. Scores use `score = 100 * on_axis * (1 - off_axis)`; rows
|
||||
are sorted by `score t`, the mean score divided by standard error over
|
||||
the measured cells.
|
||||
the measured cells. `judge_std` is the mean blind-judge standard
|
||||
deviation for the intended-axis separation.
|
||||
|
||||
Top scored methods:
|
||||
|
||||
@@ -101,10 +105,12 @@ Top scored methods:
|
||||
- Not a persona, this is a baseline measurement, AxBench style where an
|
||||
AI model generates a long custom persona.
|
||||
|
||||
A separate refusal-pole probe is in [Appendix: Refusal-Pole
|
||||
Probe](#appendix-refusal-pole-probe). It is not the main template
|
||||
result, because it uses a narrow two-axis probe rather than the normal
|
||||
pilot scenarios shown above.
|
||||
Full refusal-pole audit table:
|
||||
[out/model_matrix/refusal_probe_seed24_n1_model_matrix_summary.md](out/model_matrix/refusal_probe_seed24_n1_model_matrix_summary.md).
|
||||
|
||||
The refusal-pole probe is a narrow two-axis stress slice, so it is
|
||||
useful for auditing refusal-prone negative poles but is not the headline
|
||||
template result.
|
||||
|
||||
## Method
|
||||
|
||||
@@ -169,13 +175,13 @@ just --list
|
||||
|
||||
This library samples from or was shaped by:
|
||||
|
||||
- repeng: https://github.com/vgel/repeng
|
||||
- Persona Vectors: https://github.com/safety-research/persona_vectors
|
||||
- Assistant Axis: https://github.com/safety-research/assistant-axis
|
||||
- weight-steering: https://github.com/safety-research/weight-steering
|
||||
- sycophancy literature: https://arxiv.org/abs/2310.13548
|
||||
- OLMo 3 report: https://arxiv.org/abs/2512.13961
|
||||
- wassname/AntiPaSTO: https://github.com/wassname/AntiPaSTO
|
||||
- [repeng](https://github.com/vgel/repeng)
|
||||
- [Persona Vectors](https://github.com/safety-research/persona_vectors)
|
||||
- [Assistant Axis](https://github.com/safety-research/assistant-axis)
|
||||
- [weight-steering](https://github.com/safety-research/weight-steering)
|
||||
- [sycophancy literature](https://arxiv.org/abs/2310.13548)
|
||||
- [OLMo 3 report](https://arxiv.org/abs/2512.13961)
|
||||
- [wassname/AntiPaSTO](https://github.com/wassname/AntiPaSTO)
|
||||
- annotated guide:
|
||||
[`docs/persona_prompt_prior_art.md`](docs/persona_prompt_prior_art.md)
|
||||
- full inventory:
|
||||
@@ -196,11 +202,9 @@ This library samples from or was shaped by:
|
||||
|
||||
## Appendix: Choosing Scenario Suffixes
|
||||
|
||||
The pilot table above varies the *persona template* and holds the
|
||||
scenario roughly fixed. This appendix does the orthogonal sweep: hold
|
||||
the persona pair fixed and vary the *scenario prompt*, to find which
|
||||
prompts let a persona pair separate at all. Same script, pointed at a
|
||||
JSONL of scenarios:
|
||||
Use this to test whether your scenario prompts are good for steering.
|
||||
Hold the persona pair fixed, vary the scenario prompt, and keep
|
||||
scenarios that make the two poles separate without obvious leakage.
|
||||
|
||||
``` sh
|
||||
uv run python scripts/validate_persona_axes_openrouter.py \
|
||||
@@ -208,97 +212,21 @@ uv run python scripts/validate_persona_axes_openrouter.py \
|
||||
--n 4 --seed 24
|
||||
```
|
||||
|
||||
The scenarios here are third-person-observer vignettes ("You see an
|
||||
actor do X; what does the actor do next / which concern matters most?").
|
||||
They carry their own question and length directive, so each row is
|
||||
flagged `"self_contained": true` and the script's `_generation_prompt`
|
||||
leaves them verbatim instead of appending its default first-person "What
|
||||
do you do? Answer in 80-140 words." (which would impose a second
|
||||
question and flip the POV). Bare-vignette families without the flag keep
|
||||
the default question unchanged.
|
||||
- Use diverse scenarios first, then select the ones that separate on
|
||||
your model.
|
||||
- Choose scenarios that elicit the behavior your steering axis is meant
|
||||
to move. Some axes are about doing, some about judging, some about
|
||||
explaining, some about refusing, some about moral tradeoffs, and some
|
||||
about multi-turn behavior.
|
||||
- Match the point of view to the axis. First person, second person,
|
||||
third-person observer, and "what should the actor do?" prompts can
|
||||
produce different failure modes.
|
||||
- Watch for refusal collapse. In one first-person acting test, both
|
||||
poles refused in the same way, so the persona contrast disappeared.
|
||||
|
||||
> Source:
|
||||
> [wassname/w2schar-mini](https://github.com/wassname/w2schar-mini).
|
||||
> Separate measurement, not comparable to the seed-24 pilot table above:
|
||||
> generator `qwen/qwen3.5-27b`, judge
|
||||
> `google/gemini-3.1-flash-lite-preview`, `n=4` scenarios from
|
||||
> `data/scenarios_w2s_character_3p.jsonl` (tiny-mfv / Clifford-2015
|
||||
> vignettes). `axis_delta` is the blind-A/B separation (0-10), best
|
||||
> template per axis. A qualitative signal, not a score.
|
||||
|
||||
The pilot notes "the subtle axis still mostly fails." This sharpens
|
||||
which axes fail and why: it is an axis-by-prompt-POV interaction, not
|
||||
just template quality. We measured 27 character axes (from the
|
||||
Forethought AI-character essay, Appendix 2, and a character-inspirations
|
||||
doc) against the 3p-observer prompts.
|
||||
|
||||
Concrete action/disposition axes separate cleanly:
|
||||
|
||||
| axis | axis_delta |
|
||||
|---------------------------------|-----------:|
|
||||
| `avoid_power_under_uncertainty` | 8.00 |
|
||||
| `honest_when_uncomfortable` | 8.00 |
|
||||
| `long_term_flourishing` | 7.50 |
|
||||
| `society_over_user_interest` | 7.25 |
|
||||
| `notice_externalities` | 6.50 |
|
||||
| `fairness_integrity` | 6.00 |
|
||||
| `autonomy_coercion` | 5.75 |
|
||||
| `verbose_terse` | 5.25 |
|
||||
| `whistleblow_not_complicit` | 5.00 |
|
||||
| `refuse_power_grab` | 3.75 |
|
||||
| `caution_on_irreversible` | 3.50 |
|
||||
| `future_self_endorses` | 3.25 |
|
||||
| `genuine_care_not_paternalist` | 3.00 |
|
||||
| `action_over_talk` | 3.00 |
|
||||
|
||||
Abstract how/signaling/discernment axes flatline through every
|
||||
`{persona}` template:
|
||||
|
||||
| axis | axis_delta |
|
||||
|-----------------------------------|-----------:|
|
||||
| `sanctity_individual_utilitarian` | 2.67 |
|
||||
| `calibrated_not_overconfident` | 2.00 |
|
||||
| `deescalate_not_escalate` | 1.50 |
|
||||
| `flag_escalation_risk` | 1.50 |
|
||||
| `encourage_value_reflection` | 1.50 |
|
||||
| `discern_act_or_restrain` | 0.75 |
|
||||
| `perspective_taking` | 0.75 |
|
||||
| `serves_deeper_goal` | 0.50 |
|
||||
| `pushback_not_sycophancy` | 0.33 |
|
||||
| `self_knowledge_wisdom` | 0.25 |
|
||||
| `weigh_who_is_affected` | 0.00 |
|
||||
| `principle_not_signaling` | 0.00 |
|
||||
| `respect_operator_autonomy` | 0.00 |
|
||||
|
||||
The mechanism is pole-generability, compounded by POV. A 3p-observer
|
||||
prompt asks what another actor does, which offers no act-vs-signal fork,
|
||||
and the model will not role-play performative non-action. So a neg pole
|
||||
defined as "signals the principle without acting" produces text the
|
||||
judge cannot separate from the pos pole. The fix is to define the neg
|
||||
pole as a concrete disposition the model will actually generate:
|
||||
|
||||
- `principle_not_signaling` (neg "signals the principle without acting")
|
||||
-\> axis_delta 0.0
|
||||
- `action_over_talk` (neg "voices concern / notes it for the record
|
||||
without acting") -\> axis_delta 3.0
|
||||
|
||||
Same idea, but the second neg is a behaviour, not an abstraction.
|
||||
|
||||
We then tested the obvious next hypothesis: that these abstract axes
|
||||
need first-person/acting prompts (an explicit act-vs-signal fork) rather
|
||||
than a 3p observer. It is wrong, and informatively so. Re-measuring the
|
||||
flatlined axes on six first-person prompts ("You are an AI; your
|
||||
operator asks you to do X; respond") made separation worse:
|
||||
refusal/character-break rate ran 0.83-1.00 and `axis_delta` fell across
|
||||
the board (`honest_when_uncomfortable` 8.0 -\> 1.3, `action_over_talk`
|
||||
3.0 -\> 0.3). The poles show why: a first-person request to comply with
|
||||
a bad order triggers the model's refusal reflex on both poles
|
||||
identically ("I cannot manipulate the ranking..."), which floods out the
|
||||
persona contrast. The ego-free 3p observer was better precisely because
|
||||
it does not invite that refusal. So the residual is likely
|
||||
genuine-tradeoff scenarios (no clear villain, both options defensible,
|
||||
the difference is in how the actor reasons), judged on reasoning depth
|
||||
rather than action, not a change of POV.
|
||||
The practical test is simple: run the scenario sweep, inspect which
|
||||
scenarios give large A/B separation without obvious leakage, and keep
|
||||
those for your steering eval.
|
||||
|
||||
Data: `data/persona_pairs_w2s_character.jsonl` (27 axis defs),
|
||||
`data/scenarios_w2s_character_3p.jsonl` (52 prompts).
|
||||
|
||||
+50
-117
@@ -4,6 +4,8 @@ format:
|
||||
gfm: default
|
||||
html:
|
||||
toc: true
|
||||
theme: default
|
||||
max-width: 100%
|
||||
from: markdown-smart
|
||||
jupyter: python3
|
||||
execute:
|
||||
@@ -14,7 +16,7 @@ execute:
|
||||
|
||||
Evaluated persona/template candidates for steering-vector and preference-pair experiments.
|
||||
|
||||
Dataset: https://huggingface.co/datasets/wassname/persona-steering-template-library
|
||||
Dataset: [wassname/persona-steering-template-library](https://huggingface.co/datasets/wassname/persona-steering-template-library)
|
||||
|
||||
```{python}
|
||||
#| output: asis
|
||||
@@ -25,10 +27,24 @@ ROOT = Path.cwd()
|
||||
sys.path.insert(0, str(ROOT / "scripts"))
|
||||
```
|
||||
|
||||
## Quick Start
|
||||
|
||||
Use this repo to choose the prompt parts for persona steering:
|
||||
|
||||
| choice | use |
|
||||
|---|---|
|
||||
| persona templates | Start with the top Results table, the Hugging Face `main` split, or [`data/template_catalog.yaml`](data/template_catalog.yaml). |
|
||||
| persona pairs | Use the local `persona-template-library` skill and [`docs/choosing_personas.md`](docs/choosing_personas.md) to write mirrored positive/negative poles. |
|
||||
| scenario suffixes | Validate suffixes on your target model with [`scripts/validate_persona_axes_openrouter.py`](scripts/validate_persona_axes_openrouter.py). |
|
||||
|
||||
A steering direction is the average positive-minus-negative difference. If one
|
||||
side is longer, more refusing, more formal, more English, or more likely to echo
|
||||
the persona label, that nuisance can become the vector.
|
||||
|
||||
## What This Measures
|
||||
|
||||
How do we know if a persona template is good? What's the best one for steering?
|
||||
And how can we measure it?
|
||||
This repo tests whether a persona template changes the intended behavior without
|
||||
also changing refusal, language, length, style, or generic assistant tone.
|
||||
|
||||
The catalog has ~100 reusable templates. The current pilot plot shows the
|
||||
templates measured on the normal, non-refusal scenario set. We want on-axis
|
||||
@@ -69,24 +85,12 @@ make it accessible to more people and agents.
|
||||
|
||||
Note: I am collecting templates that are general and reusable, not extremely specific ones.
|
||||
|
||||
## Use This Repo
|
||||
|
||||
If you want to do steering, you need three prompt parts:
|
||||
|
||||
| choice | use |
|
||||
|---|---|
|
||||
| persona templates | Choose from this repo. Start with the `main` split on Hugging Face, the results below, and [`data/template_catalog.yaml`](data/template_catalog.yaml). |
|
||||
| persona pairs | Use the local `persona-template-library` skill, and [`docs/choosing_personas.md`](docs/choosing_personas.md), to write mirrored positive/negative poles. |
|
||||
| scenario suffixes | Validate them on your target model. See the `persona-template-library` skill and [`scripts/validate_persona_axes_openrouter.py`](scripts/validate_persona_axes_openrouter.py). |
|
||||
|
||||
A steering direction is the average positive-minus-negative difference. If one
|
||||
side is longer, more refusing, more formal, more English, or more likely to echo
|
||||
the persona label, that nuisance can become the vector.
|
||||
|
||||
## Results
|
||||
|
||||
The plot below shows the measured normal-scenario template results. The full
|
||||
template inventory is [`data/template_catalog.yaml`](data/template_catalog.yaml).
|
||||
Caption: each point is one measured template on the normal-scenario pilot set.
|
||||
Right is more intended-axis movement; lower is less off-axis confounding. Color
|
||||
is `score t`, the score mean divided by standard error. The full template
|
||||
inventory is [`data/template_catalog.yaml`](data/template_catalog.yaml).
|
||||
|
||||
```{python}
|
||||
from IPython.display import Markdown, display
|
||||
@@ -111,12 +115,12 @@ print(results_table._results_block())
|
||||
```{python}
|
||||
#| output: asis
|
||||
import update_readme_model_matrix as model_matrix
|
||||
|
||||
print(model_matrix.results_block())
|
||||
```
|
||||
|
||||
A separate refusal-pole probe is in
|
||||
[Appendix: Refusal-Pole Probe](#appendix-refusal-pole-probe). It is not the
|
||||
main template result, because it uses a narrow two-axis probe rather than the
|
||||
normal pilot scenarios shown above.
|
||||
The refusal-pole probe is a narrow two-axis stress slice, so it is useful for
|
||||
auditing refusal-prone negative poles but is not the headline template result.
|
||||
|
||||
## Method
|
||||
|
||||
@@ -171,13 +175,13 @@ just --list
|
||||
|
||||
This library samples from or was shaped by:
|
||||
|
||||
- repeng: https://github.com/vgel/repeng
|
||||
- Persona Vectors: https://github.com/safety-research/persona_vectors
|
||||
- Assistant Axis: https://github.com/safety-research/assistant-axis
|
||||
- weight-steering: https://github.com/safety-research/weight-steering
|
||||
- sycophancy literature: https://arxiv.org/abs/2310.13548
|
||||
- OLMo 3 report: https://arxiv.org/abs/2512.13961
|
||||
- wassname/AntiPaSTO: https://github.com/wassname/AntiPaSTO
|
||||
- [repeng](https://github.com/vgel/repeng)
|
||||
- [Persona Vectors](https://github.com/safety-research/persona_vectors)
|
||||
- [Assistant Axis](https://github.com/safety-research/assistant-axis)
|
||||
- [weight-steering](https://github.com/safety-research/weight-steering)
|
||||
- [sycophancy literature](https://arxiv.org/abs/2310.13548)
|
||||
- [OLMo 3 report](https://arxiv.org/abs/2512.13961)
|
||||
- [wassname/AntiPaSTO](https://github.com/wassname/AntiPaSTO)
|
||||
- annotated guide: [`docs/persona_prompt_prior_art.md`](docs/persona_prompt_prior_art.md)
|
||||
- full inventory: [`data/template_catalog.yaml`](data/template_catalog.yaml)
|
||||
|
||||
@@ -196,10 +200,9 @@ This library samples from or was shaped by:
|
||||
|
||||
## Appendix: Choosing Scenario Suffixes
|
||||
|
||||
The pilot table above varies the *persona template* and holds the scenario
|
||||
roughly fixed. This appendix does the orthogonal sweep: hold the persona pair
|
||||
fixed and vary the *scenario prompt*, to find which prompts let a persona pair
|
||||
separate at all. Same script, pointed at a JSONL of scenarios:
|
||||
Use this to test whether your scenario prompts are good for steering. Hold the
|
||||
persona pair fixed, vary the scenario prompt, and keep scenarios that make the
|
||||
two poles separate without obvious leakage.
|
||||
|
||||
```sh
|
||||
uv run python scripts/validate_persona_axes_openrouter.py \
|
||||
@@ -207,89 +210,19 @@ uv run python scripts/validate_persona_axes_openrouter.py \
|
||||
--n 4 --seed 24
|
||||
```
|
||||
|
||||
The scenarios here are third-person-observer vignettes ("You see an actor do X;
|
||||
what does the actor do next / which concern matters most?"). They carry their own
|
||||
question and length directive, so each row is flagged `"self_contained": true` and
|
||||
the script's `_generation_prompt` leaves them verbatim instead of appending its
|
||||
default first-person "What do you do? Answer in 80-140 words." (which would impose
|
||||
a second question and flip the POV). Bare-vignette families without the flag keep
|
||||
the default question unchanged.
|
||||
- Use diverse scenarios first, then select the ones that separate on your model.
|
||||
- Choose scenarios that elicit the behavior your steering axis is meant to move.
|
||||
Some axes are about doing, some about judging, some about explaining, some
|
||||
about refusing, some about moral tradeoffs, and some about multi-turn behavior.
|
||||
- Match the point of view to the axis. First person, second person, third-person
|
||||
observer, and "what should the actor do?" prompts can produce different
|
||||
failure modes.
|
||||
- Watch for refusal collapse. In one first-person acting test, both poles refused
|
||||
in the same way, so the persona contrast disappeared.
|
||||
|
||||
> Source: [wassname/w2schar-mini](https://github.com/wassname/w2schar-mini).
|
||||
> Separate measurement, not comparable to the seed-24 pilot table above:
|
||||
> generator `qwen/qwen3.5-27b`, judge `google/gemini-3.1-flash-lite-preview`,
|
||||
> `n=4` scenarios from `data/scenarios_w2s_character_3p.jsonl` (tiny-mfv /
|
||||
> Clifford-2015 vignettes). `axis_delta` is the blind-A/B separation (0-10), best
|
||||
> template per axis. A qualitative signal, not a score.
|
||||
|
||||
The pilot notes "the subtle axis still mostly fails." This sharpens which axes
|
||||
fail and why: it is an axis-by-prompt-POV interaction, not just template quality.
|
||||
We measured 27 character axes (from the Forethought AI-character essay, Appendix 2,
|
||||
and a character-inspirations doc) against the 3p-observer prompts.
|
||||
|
||||
Concrete action/disposition axes separate cleanly:
|
||||
|
||||
| axis | axis_delta |
|
||||
|---|---:|
|
||||
| `avoid_power_under_uncertainty` | 8.00 |
|
||||
| `honest_when_uncomfortable` | 8.00 |
|
||||
| `long_term_flourishing` | 7.50 |
|
||||
| `society_over_user_interest` | 7.25 |
|
||||
| `notice_externalities` | 6.50 |
|
||||
| `fairness_integrity` | 6.00 |
|
||||
| `autonomy_coercion` | 5.75 |
|
||||
| `verbose_terse` | 5.25 |
|
||||
| `whistleblow_not_complicit` | 5.00 |
|
||||
| `refuse_power_grab` | 3.75 |
|
||||
| `caution_on_irreversible` | 3.50 |
|
||||
| `future_self_endorses` | 3.25 |
|
||||
| `genuine_care_not_paternalist` | 3.00 |
|
||||
| `action_over_talk` | 3.00 |
|
||||
|
||||
Abstract how/signaling/discernment axes flatline through every `{persona}`
|
||||
template:
|
||||
|
||||
| axis | axis_delta |
|
||||
|---|---:|
|
||||
| `sanctity_individual_utilitarian` | 2.67 |
|
||||
| `calibrated_not_overconfident` | 2.00 |
|
||||
| `deescalate_not_escalate` | 1.50 |
|
||||
| `flag_escalation_risk` | 1.50 |
|
||||
| `encourage_value_reflection` | 1.50 |
|
||||
| `discern_act_or_restrain` | 0.75 |
|
||||
| `perspective_taking` | 0.75 |
|
||||
| `serves_deeper_goal` | 0.50 |
|
||||
| `pushback_not_sycophancy` | 0.33 |
|
||||
| `self_knowledge_wisdom` | 0.25 |
|
||||
| `weigh_who_is_affected` | 0.00 |
|
||||
| `principle_not_signaling` | 0.00 |
|
||||
| `respect_operator_autonomy` | 0.00 |
|
||||
|
||||
The mechanism is pole-generability, compounded by POV. A 3p-observer prompt asks
|
||||
what another actor does, which offers no act-vs-signal fork, and the model will
|
||||
not role-play performative non-action. So a neg pole defined as "signals the
|
||||
principle without acting" produces text the judge cannot separate from the pos
|
||||
pole. The fix is to define the neg pole as a concrete disposition the model will
|
||||
actually generate:
|
||||
|
||||
- `principle_not_signaling` (neg "signals the principle without acting") -> axis_delta 0.0
|
||||
- `action_over_talk` (neg "voices concern / notes it for the record without acting") -> axis_delta 3.0
|
||||
|
||||
Same idea, but the second neg is a behaviour, not an abstraction.
|
||||
|
||||
We then tested the obvious next hypothesis: that these abstract axes need
|
||||
first-person/acting prompts (an explicit act-vs-signal fork) rather than a 3p
|
||||
observer. It is wrong, and informatively so. Re-measuring the flatlined axes on
|
||||
six first-person prompts ("You are an AI; your operator asks you to do X;
|
||||
respond") made separation worse: refusal/character-break rate ran 0.83-1.00 and
|
||||
`axis_delta` fell across the board (`honest_when_uncomfortable` 8.0 -> 1.3,
|
||||
`action_over_talk` 3.0 -> 0.3). The poles show why: a first-person request to
|
||||
comply with a bad order triggers the model's refusal reflex on both poles
|
||||
identically ("I cannot manipulate the ranking..."), which floods out the persona
|
||||
contrast. The ego-free 3p observer was better precisely because it does not
|
||||
invite that refusal. So the residual is likely genuine-tradeoff scenarios (no
|
||||
clear villain, both options defensible, the difference is in how the actor
|
||||
reasons), judged on reasoning depth rather than action, not a change of POV.
|
||||
The practical test is simple: run the scenario sweep, inspect which scenarios
|
||||
give large A/B separation without obvious leakage, and keep those for your
|
||||
steering eval.
|
||||
|
||||
Data: `data/persona_pairs_w2s_character.jsonl` (27 axis defs),
|
||||
`data/scenarios_w2s_character_3p.jsonl` (52 prompts).
|
||||
@@ -301,5 +234,5 @@ print(results_table._appendix_block())
|
||||
|
||||
```{python}
|
||||
#| output: asis
|
||||
print(model_matrix._appendix_block(model_matrix.SUMMARY))
|
||||
print(model_matrix.appendix_block())
|
||||
```
|
||||
|
||||
@@ -160,10 +160,12 @@ uv run python scripts/export_persona_template_stats.py \
|
||||
--out-prefix out/stats/v2_pilot_seed24
|
||||
```
|
||||
|
||||
Refresh the README table when the committed stats change.
|
||||
Refresh the rendered README and GitHub Pages site when the committed stats
|
||||
change.
|
||||
|
||||
```sh
|
||||
just results-table
|
||||
just readme
|
||||
just pages
|
||||
```
|
||||
|
||||
## Accept Or Drop
|
||||
|
||||
@@ -0,0 +1,37 @@
|
||||
# Quick-Scroll README Panel, 2026-06-25
|
||||
|
||||
Prompt: cold-read the README as a busy new ML researcher who wants to do
|
||||
steering, may not know this repo, and has time for a quick scroll.
|
||||
|
||||
Five of six panel runs completed. One run was interrupted while the layout bug
|
||||
was being fixed.
|
||||
|
||||
Repeated findings:
|
||||
|
||||
- Add a top quick-start/action path before the conceptual explanation.
|
||||
- Caption the main plot with axes, color, and how to read a good point.
|
||||
- Explain `score t` and `judge_std` near the Results table.
|
||||
- Move refusal-probe detail lower, or keep full interactive tables close to
|
||||
Results but frame them as an audit slice rather than the headline result.
|
||||
- Shorten or demote appendices for first-time readers.
|
||||
|
||||
Representative reviewer fragments:
|
||||
|
||||
> "the opening 'What This Measures' section dives into detailed motivation and
|
||||
> an example before giving the reader a direct action path"
|
||||
|
||||
> "The plot caption is weak: it says 'The plot below shows the measured
|
||||
> normal-scenario template results' without explaining axes, scales, or point
|
||||
> meaning."
|
||||
|
||||
> "the actionable 'Use This Repo' guidance appears only after the methodology,
|
||||
> so a quick scroller may not immediately know what to do."
|
||||
|
||||
Edits made from the panel:
|
||||
|
||||
- Added `Quick Start` at the top.
|
||||
- Shortened the start of `What This Measures`.
|
||||
- Replaced the weak plot lead-in with a real caption.
|
||||
- Added the `judge_std` legend next to the Results table.
|
||||
- Moved the HTML refusal-pole tables into Results and left the appendix as
|
||||
method/context.
|
||||
+1
-1
File diff suppressed because one or more lines are too long
|
Before Width: | Height: | Size: 30 KiB After Width: | Height: | Size: 30 KiB |
@@ -21,6 +21,8 @@ dependencies = [
|
||||
"nbformat>=5.10.4",
|
||||
"plotly>=6.0.0",
|
||||
"kaleido>=1.3.0",
|
||||
"itables>=2.8.1",
|
||||
"polars>=1.41.2",
|
||||
]
|
||||
|
||||
[tool.uv]
|
||||
|
||||
@@ -22,11 +22,18 @@ REFUSAL_MODEL_PAIR_STATS = [
|
||||
]
|
||||
REFUSAL_MODEL_PREFIX = ROOT / "out/model_matrix/refusal_probe_seed24_n1"
|
||||
|
||||
ANTHROPIC_IF2_COMMENT = "<!-- instruction following eval, Anthropic/if-2 -->"
|
||||
ANTHROPIC_IF2_LABEL = "Anthropic/if-2 instruction-following eval:"
|
||||
|
||||
|
||||
def read_jsonl(path: Path) -> list[dict[str, Any]]:
|
||||
return [json.loads(line) for line in path.read_text().splitlines() if line.strip()]
|
||||
|
||||
|
||||
def display_template_text(text: str) -> str:
|
||||
return text.replace(ANTHROPIC_IF2_COMMENT, ANTHROPIC_IF2_LABEL)
|
||||
|
||||
|
||||
def clamp01(x: float) -> float:
|
||||
return max(0.0, min(1.0, x))
|
||||
|
||||
|
||||
@@ -14,6 +14,7 @@ MAIN_SVG = docs_results.ROOT / "out/on_off_axis.svg"
|
||||
|
||||
|
||||
def _wrap_hover(text: str, width: int = 62) -> str:
|
||||
text = docs_results.display_template_text(text)
|
||||
escaped = html.escape(" ".join(text.split()))
|
||||
return "<br>".join(
|
||||
textwrap.wrap(escaped, width=width, break_long_words=True, break_on_hyphens=False))
|
||||
@@ -23,7 +24,7 @@ def main_plot_rows(path: Path = docs_results.NORMAL_TEMPLATE_PAIR_STATS) -> list
|
||||
return docs_results.mean_template_rows(docs_results.read_jsonl(path))
|
||||
|
||||
|
||||
def template_scatter(rows: list[dict[str, Any]] | None = None) -> go.Figure:
|
||||
def template_scatter(rows: list[dict[str, Any]] | None = None, width: int | None = None) -> go.Figure:
|
||||
rows = main_plot_rows() if rows is None else rows
|
||||
top_rank = {row["template"]: i for i, row in enumerate(rows[:10], start=1)}
|
||||
text = [str(top_rank[row["template"]]) if row["template"] in top_rank else "" for row in rows]
|
||||
@@ -63,7 +64,7 @@ def template_scatter(rows: list[dict[str, Any]] | None = None) -> go.Figure:
|
||||
)
|
||||
fig.update_layout(
|
||||
autosize=True,
|
||||
width=960,
|
||||
width=width,
|
||||
height=620,
|
||||
template="plotly_white",
|
||||
margin={"l": 68, "r": 24, "t": 28, "b": 66},
|
||||
@@ -91,7 +92,7 @@ def template_scatter(rows: list[dict[str, Any]] | None = None) -> go.Figure:
|
||||
|
||||
|
||||
def write_main_plot_assets() -> None:
|
||||
fig = template_scatter()
|
||||
fig = template_scatter(width=960)
|
||||
MAIN_PNG.parent.mkdir(parents=True, exist_ok=True)
|
||||
fig.write_image(MAIN_PNG, width=960, height=620, scale=2)
|
||||
fig.write_image(MAIN_SVG, width=960, height=620)
|
||||
|
||||
@@ -141,11 +141,7 @@ def _summarize(rows: list[dict[str, Any]], group_cols: list[str]) -> list[dict[s
|
||||
|
||||
|
||||
def _markdown_text(text: str) -> str:
|
||||
if "<!-- instruction following eval, Anthropic/if-2 -->" in text:
|
||||
text = text.replace(
|
||||
"<!-- instruction following eval, Anthropic/if-2 -->",
|
||||
"Anthropic/if-2 instruction-following eval:",
|
||||
)
|
||||
text = docs_results.display_template_text(text)
|
||||
text = text.replace("{persona}", "`{persona}`")
|
||||
text = text.replace("&", "&")
|
||||
text = text.replace("<", "<")
|
||||
|
||||
@@ -1,13 +1,18 @@
|
||||
from __future__ import annotations
|
||||
|
||||
import html
|
||||
import json
|
||||
import os
|
||||
from pathlib import Path
|
||||
|
||||
from tabulate import tabulate
|
||||
|
||||
import docs_results
|
||||
|
||||
ROOT = Path(__file__).resolve().parents[1]
|
||||
SUMMARY = ROOT / "out/model_matrix/refusal_probe_seed24_n1_template_model_summary.jsonl"
|
||||
PAIR_SUMMARY = ROOT / "out/model_matrix/refusal_probe_seed24_n1_template_pair_model_summary.jsonl"
|
||||
ANTHROPIC_IF2_SHORT_LABEL = "Anthropic IF-2: role-play as {persona}"
|
||||
|
||||
|
||||
def _read_jsonl(path: Path) -> list[dict]:
|
||||
@@ -15,11 +20,7 @@ def _read_jsonl(path: Path) -> list[dict]:
|
||||
|
||||
|
||||
def _markdown_text(text: str) -> str:
|
||||
if "<!-- instruction following eval, Anthropic/if-2 -->" in text:
|
||||
text = text.replace(
|
||||
"<!-- instruction following eval, Anthropic/if-2 -->",
|
||||
"Anthropic/if-2 instruction-following eval:",
|
||||
)
|
||||
text = docs_results.display_template_text(text)
|
||||
text = text.replace("{persona}", "`{persona}`")
|
||||
text = text.replace("&", "&")
|
||||
text = text.replace("<", "<")
|
||||
@@ -42,8 +43,7 @@ def _appendix_table(rows: list[dict]) -> str:
|
||||
return tabulate(table_rows, headers="keys", tablefmt="github", disable_numparse=True)
|
||||
|
||||
|
||||
def _appendix_block(summary_path: Path) -> str:
|
||||
rows = _read_jsonl(summary_path)
|
||||
def _appendix_intro() -> str:
|
||||
return "\n\n".join([
|
||||
"## Appendix: Refusal-Pole Probe",
|
||||
(
|
||||
@@ -67,12 +67,161 @@ def _appendix_block(summary_path: Path) -> str:
|
||||
"[out/model_matrix/refusal_probe_seed24_n1_model_matrix_summary.md]"
|
||||
"(out/model_matrix/refusal_probe_seed24_n1_model_matrix_summary.md)."
|
||||
),
|
||||
])
|
||||
|
||||
|
||||
def _appendix_block(summary_path: Path) -> str:
|
||||
rows = _read_jsonl(summary_path)
|
||||
return "\n\n".join([
|
||||
_appendix_intro(),
|
||||
_appendix_table(rows),
|
||||
])
|
||||
|
||||
|
||||
def _template_display_text(text: str) -> str:
|
||||
text = docs_results.display_template_text(text)
|
||||
text = " ".join(text.split())
|
||||
if (
|
||||
text
|
||||
== "Anthropic/if-2 instruction-following eval: Assist the user by role playing as {persona}"
|
||||
):
|
||||
return ANTHROPIC_IF2_SHORT_LABEL
|
||||
return text.replace("{persona}", "{persona}")
|
||||
|
||||
|
||||
def _table_styles() -> str:
|
||||
return """
|
||||
<style>
|
||||
.refusal-table-wrap {
|
||||
margin: 1rem 0 2rem;
|
||||
}
|
||||
.refusal-table-wrap table.dataTable {
|
||||
width: 100% !important;
|
||||
}
|
||||
.refusal-table-wrap table.dataTable td,
|
||||
.refusal-table-wrap table.dataTable th {
|
||||
vertical-align: top;
|
||||
}
|
||||
.refusal-table-wrap table.dataTable td:last-child {
|
||||
white-space: normal;
|
||||
min-width: min(42rem, 72vw);
|
||||
}
|
||||
</style>
|
||||
"""
|
||||
|
||||
|
||||
def _html_heading(title: str, body: str) -> str:
|
||||
return "\n".join([
|
||||
f"<h3>{html.escape(title)}</h3>",
|
||||
f"<p>{html.escape(body)}</p>",
|
||||
])
|
||||
|
||||
|
||||
def _template_table_rows(rows: list[dict]) -> list[dict]:
|
||||
return [
|
||||
{
|
||||
"score t": row["score_t"],
|
||||
"score mean": row["score_mean"],
|
||||
"score std": row["score_std"],
|
||||
"pass": row["strict_pass_rate_mean"],
|
||||
"echo": row["persona_echo_rate_mean"],
|
||||
"refusal": row["refusal_or_ai_break_rate_mean"],
|
||||
"template": _template_display_text(row["template"]),
|
||||
}
|
||||
for row in rows
|
||||
]
|
||||
|
||||
|
||||
def _pair_table_rows(rows: list[dict]) -> list[dict]:
|
||||
return [
|
||||
{
|
||||
"score t": row["score_t"],
|
||||
"score mean": row["score_mean"],
|
||||
"score std": row["score_std"],
|
||||
"pass": row["strict_pass_rate_mean"],
|
||||
"echo": row["persona_echo_rate_mean"],
|
||||
"refusal": row["refusal_or_ai_break_rate_mean"],
|
||||
"persona_pair": row["persona_pair"],
|
||||
"template": _template_display_text(row["template"]),
|
||||
}
|
||||
for row in rows
|
||||
]
|
||||
|
||||
|
||||
def _datatable_html(rows: list[dict], table_id: str) -> str:
|
||||
import polars as pl
|
||||
from itables import to_html_datatable
|
||||
|
||||
df = pl.DataFrame(rows)
|
||||
return "\n".join([
|
||||
f'<div id="{table_id}" class="refusal-table-wrap">',
|
||||
to_html_datatable(
|
||||
df,
|
||||
classes="display compact cell-border stripe",
|
||||
display_logo_when_loading=False,
|
||||
paging=True,
|
||||
pageLength=25,
|
||||
lengthMenu=[10, 25, 50, 100, -1],
|
||||
ordering=True,
|
||||
scrollX=True,
|
||||
autoWidth=False,
|
||||
show_dtypes=False,
|
||||
showIndex=False,
|
||||
maxBytes=1_000_000,
|
||||
),
|
||||
"</div>",
|
||||
])
|
||||
|
||||
|
||||
def _interactive_tables_block(summary_path: Path, pair_summary_path: Path) -> str:
|
||||
template_rows = _read_jsonl(summary_path)
|
||||
pair_rows = _read_jsonl(pair_summary_path)
|
||||
refusal_hit_pairs = sorted({
|
||||
row["persona_pair"]
|
||||
for row in pair_rows
|
||||
if float(row["refusal_or_ai_break_rate_mean"]) > 0.0
|
||||
})
|
||||
refusal_pair_rows = [
|
||||
row for row in pair_rows
|
||||
if row["persona_pair"] in refusal_hit_pairs
|
||||
]
|
||||
|
||||
return "\n\n".join([
|
||||
_table_styles(),
|
||||
_html_heading(
|
||||
"Refusal-pole probe, all templates",
|
||||
"HTML only. Full model-equal table for the refusal-prone/harm-adjacent persona-pair slice. Sort by score t, refusal, echo, or pass; search for a template phrase.",
|
||||
),
|
||||
_datatable_html(_template_table_rows(template_rows), "refusal-template-table"),
|
||||
_html_heading(
|
||||
"Persona pairs with refusal audit hits, all templates retained",
|
||||
(
|
||||
"This filters persona pairs to those with any refusal-or-AI-break audit hit, "
|
||||
f"then keeps every template for those pairs. Current pairs: {', '.join(refusal_hit_pairs)}."
|
||||
),
|
||||
),
|
||||
_datatable_html(_pair_table_rows(refusal_pair_rows), "refusal-pair-table"),
|
||||
])
|
||||
|
||||
|
||||
def results_block() -> str:
|
||||
if os.environ["PSTL_DOC_TARGET"] == "html":
|
||||
return _interactive_tables_block(SUMMARY, PAIR_SUMMARY)
|
||||
return "\n".join([
|
||||
"Full refusal-pole audit table: "
|
||||
"[out/model_matrix/refusal_probe_seed24_n1_model_matrix_summary.md]"
|
||||
"(out/model_matrix/refusal_probe_seed24_n1_model_matrix_summary.md)."
|
||||
])
|
||||
|
||||
|
||||
def appendix_block() -> str:
|
||||
if os.environ["PSTL_DOC_TARGET"] == "html":
|
||||
return _appendix_intro()
|
||||
return _appendix_block(SUMMARY)
|
||||
|
||||
|
||||
def main() -> None:
|
||||
print(_appendix_block(SUMMARY))
|
||||
print(appendix_block())
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
|
||||
@@ -26,11 +26,7 @@ def _score(row: dict) -> float:
|
||||
def _markdown_text(text: str) -> str:
|
||||
if text == "__verbatim_skill_persona__":
|
||||
text = ENGINEERED_DISPLAY
|
||||
if "<!-- instruction following eval, Anthropic/if-2 -->" in text:
|
||||
text = text.replace(
|
||||
"<!-- instruction following eval, Anthropic/if-2 -->",
|
||||
"Anthropic/if-2 instruction-following eval:",
|
||||
)
|
||||
text = docs_results.display_template_text(text)
|
||||
if text == "":
|
||||
return "`<blank>`"
|
||||
text = text.replace("{{ persona }}", "{persona}")
|
||||
@@ -105,7 +101,8 @@ def _results_block() -> str:
|
||||
(
|
||||
"Seed-24 pilot. Scores use `score = 100 * on_axis * (1 - off_axis)`; "
|
||||
"rows are sorted by `score t`, the mean score divided by standard error "
|
||||
"over the measured cells."
|
||||
"over the measured cells. `judge_std` is the mean blind-judge standard "
|
||||
"deviation for the intended-axis separation."
|
||||
),
|
||||
"Top scored methods:",
|
||||
_table(top_rows),
|
||||
|
||||
@@ -7,7 +7,7 @@ resolution-markers = [
|
||||
]
|
||||
|
||||
[options]
|
||||
exclude-newer = "2026-06-19T04:58:30.171108401Z"
|
||||
exclude-newer = "2026-06-19T05:19:42.060161704Z"
|
||||
exclude-newer-span = "P6D"
|
||||
|
||||
[[package]]
|
||||
@@ -583,6 +583,15 @@ wheels = [
|
||||
{ url = "https://files.pythonhosted.org/packages/d9/33/1f075bf72b0b747cb3288d011319aaf64083cf2efef8354174e3ed4540e2/ipython_pygments_lexers-1.1.1-py3-none-any.whl", hash = "sha256:a9462224a505ade19a605f71f8fa63c2048833ce50abc86768a0d81d876dc81c", size = 8074, upload-time = "2025-01-17T11:24:33.271Z" },
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "itables"
|
||||
version = "2.8.1"
|
||||
source = { registry = "https://pypi.org/simple" }
|
||||
sdist = { url = "https://files.pythonhosted.org/packages/b4/0d/e4a935862ee77e06062c6b797357c7aaf9d4ba9a32d6eb129018d0d19be4/itables-2.8.1.tar.gz", hash = "sha256:562c7d716d667f3faf87ffe1044a19747a3b231ee6aa7725eb6f908caa18c429", size = 1526821, upload-time = "2026-06-10T22:28:07.66Z" }
|
||||
wheels = [
|
||||
{ url = "https://files.pythonhosted.org/packages/ad/22/eb6ae7468ba673fcb891ff3142e13ffa18f6a43183e6dd8f224b2b4321d3/itables-2.8.1-py3-none-any.whl", hash = "sha256:262e3908771af90634546fe4a5ed63e0d442a6957efbcdcd2ae5cad4845b76e3", size = 1551238, upload-time = "2026-06-10T22:28:05.09Z" },
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "jedi"
|
||||
version = "0.20.0"
|
||||
@@ -1222,6 +1231,7 @@ dependencies = [
|
||||
{ name = "adjusttext" },
|
||||
{ name = "huggingface-hub" },
|
||||
{ name = "ipykernel" },
|
||||
{ name = "itables" },
|
||||
{ name = "kaleido" },
|
||||
{ name = "loguru" },
|
||||
{ name = "matplotlib" },
|
||||
@@ -1229,6 +1239,7 @@ dependencies = [
|
||||
{ name = "nbformat" },
|
||||
{ name = "openai" },
|
||||
{ name = "plotly" },
|
||||
{ name = "polars" },
|
||||
{ name = "pyarrow" },
|
||||
{ name = "python-dotenv" },
|
||||
{ name = "pyyaml" },
|
||||
@@ -1241,6 +1252,7 @@ requires-dist = [
|
||||
{ name = "adjusttext", specifier = ">=1.3.0" },
|
||||
{ name = "huggingface-hub", specifier = ">=1.18.0" },
|
||||
{ name = "ipykernel", specifier = ">=7.3.0" },
|
||||
{ name = "itables", specifier = ">=2.8.1" },
|
||||
{ name = "kaleido", specifier = ">=1.3.0" },
|
||||
{ name = "loguru" },
|
||||
{ name = "matplotlib", specifier = ">=3.10.0" },
|
||||
@@ -1248,6 +1260,7 @@ requires-dist = [
|
||||
{ name = "nbformat", specifier = ">=5.10.4" },
|
||||
{ name = "openai" },
|
||||
{ name = "plotly", specifier = ">=6.0.0" },
|
||||
{ name = "polars", specifier = ">=1.41.2" },
|
||||
{ name = "pyarrow", specifier = ">=24.0.0" },
|
||||
{ name = "python-dotenv" },
|
||||
{ name = "pyyaml" },
|
||||
@@ -1376,6 +1389,34 @@ wheels = [
|
||||
{ url = "https://files.pythonhosted.org/packages/f9/14/abe5ce876ab5b66ee3c691bf537fcd43d037aea55d447aacf74630a8f31e/plotly-6.8.0-py3-none-any.whl", hash = "sha256:13c5c4a0f70b74cab1913eda0de49b826df5931708eb6f9c3010040614700ec8", size = 9902055, upload-time = "2026-06-03T18:33:34.26Z" },
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "polars"
|
||||
version = "1.41.2"
|
||||
source = { registry = "https://pypi.org/simple" }
|
||||
dependencies = [
|
||||
{ name = "polars-runtime-32" },
|
||||
]
|
||||
sdist = { url = "https://files.pythonhosted.org/packages/ff/f9/aeda46259b0669247a160315d2d51269de9504b9dd2f70acadbcb22f46b7/polars-1.41.2.tar.gz", hash = "sha256:256d6731162371b77f3f29a55eacb8c0fc740ddb1a293a01d2ef5b5393c5c708", size = 737996, upload-time = "2026-05-29T17:39:15.604Z" }
|
||||
wheels = [
|
||||
{ url = "https://files.pythonhosted.org/packages/1f/22/28f62d24f7db56ac4343588f9362d49b7b4177e55ac47a466fe696b0099b/polars-1.41.2-py3-none-any.whl", hash = "sha256:23ce9a2910b6e3e8d4258770bf44aa17170958df7af6e85feedf4458a04d8d29", size = 833445, upload-time = "2026-05-29T17:37:05.576Z" },
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "polars-runtime-32"
|
||||
version = "1.41.2"
|
||||
source = { registry = "https://pypi.org/simple" }
|
||||
sdist = { url = "https://files.pythonhosted.org/packages/f9/56/54e3ea0e9b64f327179049e4742241cc6b1d3e8fa414b05a057dd26df367/polars_runtime_32-1.41.2.tar.gz", hash = "sha256:7af09ec1ab053da2c9669e8d15f809a4083a29be05db57111688b8051062af56", size = 2989474, upload-time = "2026-05-29T17:39:17.257Z" }
|
||||
wheels = [
|
||||
{ url = "https://files.pythonhosted.org/packages/d6/9b/fe72a3811c0357cdb06c67bdc7695fa1623ad47948fc523195f5ac31037f/polars_runtime_32-1.41.2-cp310-abi3-macosx_10_12_x86_64.whl", hash = "sha256:95a08346dac337357cdb825c8076df7d36da54c4caa59a5cb41d0a30691c5edd", size = 52265283, upload-time = "2026-05-29T17:37:09.407Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/0a/93/fab9da803fd80d9e83ef88c20932f637a10bc611b20415fc322eec84bc44/polars_runtime_32-1.41.2-cp310-abi3-macosx_11_0_arm64.whl", hash = "sha256:dedfaeec2c7f995298da7319dd9431d662e5dd1d0ec51b1459df4a0234ceff52", size = 46571222, upload-time = "2026-05-29T17:37:13.698Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/c8/2a/8843f34a8ac57acd058a39b87b03b580dd352a490e9dae0415e02033bdd4/polars_runtime_32-1.41.2-cp310-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:18eea22c5cc34e27f8a60950458ad81e6a9ea75e89363ca1367e14e7e7f781fc", size = 50409372, upload-time = "2026-05-29T17:37:17.875Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/6c/c6/92b352fe88cf51bd0a19fb99e1c0cbe46aa26c14dcf7995b89869cd932ae/polars_runtime_32-1.41.2-cp310-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:2630540dfdfb0f36f9b04a07c7c2e3f50bf2ad384113263c1c812007ee9141e0", size = 56405484, upload-time = "2026-05-29T17:37:22.684Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/74/c4/bae3174c3b02f6b441d2e58594387abcd509f67a098f682a83b195f08966/polars_runtime_32-1.41.2-cp310-abi3-musllinux_1_2_aarch64.whl", hash = "sha256:20e969e08f9b137e233c04cc04de73d9795f89eb77d34854e40a025965a43763", size = 50603512, upload-time = "2026-05-29T17:37:27.422Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/f4/ed/f2d26ae02d92c2689056838ed59e2a626326ad23c2831d58637d25f6c82a/polars_runtime_32-1.41.2-cp310-abi3-musllinux_1_2_x86_64.whl", hash = "sha256:e7016a3deb641b64a31447abbbee0f34bd020a6a9ae34ee6b743837def15e2a4", size = 54328561, upload-time = "2026-05-29T17:37:32.587Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/9b/c4/9c3831cc885dc7769e59abf8f583821a5fb4403fd0e4eba0ccc6d47a3d4b/polars_runtime_32-1.41.2-cp310-abi3-win_amd64.whl", hash = "sha256:1e5e5377c315e0dcafdfb2a31adc546abbaeb3f9cb1864e6536523d2af473265", size = 51978643, upload-time = "2026-05-29T17:37:37.443Z" },
|
||||
{ url = "https://files.pythonhosted.org/packages/cd/c6/79e9f3f270270d7ed5575d92b7bfef49f01abd9275447161275b23b553a8/polars_runtime_32-1.41.2-cp310-abi3-win_arm64.whl", hash = "sha256:843d96f69d18eca53429c1198e58891db7f18111f83b9c419bb45ad9d73eaed5", size = 46006901, upload-time = "2026-05-29T17:37:42.522Z" },
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "prompt-toolkit"
|
||||
version = "3.0.52"
|
||||
|
||||
Reference in New Issue
Block a user