Datasets:
v1.0.0 — initial release
Browse files- README.md +183 -0
- eval.jsonl +175 -0
- seeds.yaml +1194 -0
- training.jsonl +0 -0
README.md
ADDED
|
@@ -0,0 +1,183 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language: en
|
| 4 |
+
size_categories:
|
| 5 |
+
- 10K<n<100K
|
| 6 |
+
task_categories:
|
| 7 |
+
- text-classification
|
| 8 |
+
tags:
|
| 9 |
+
- trading
|
| 10 |
+
- intent-classification
|
| 11 |
+
- human-curated
|
| 12 |
+
- augmented
|
| 13 |
+
- english
|
| 14 |
+
---
|
| 15 |
+
|
| 16 |
+
# trade-decision-classifier-v1-dataset
|
| 17 |
+
|
| 18 |
+
Human-curated + synthetically augmented dataset for training trading-agent reply classifiers. Companion to [DoDataThings/distilbert-trade-decision-classifier-v1](https://huggingface.co/DoDataThings/distilbert-trade-decision-classifier-v1).
|
| 19 |
+
|
| 20 |
+
## Why this dataset
|
| 21 |
+
|
| 22 |
+
Trading agents that DM trade proposals to a human reviewer and accept free-form text replies need a way to convert those replies into discrete actions. There's no off-the-shelf dataset for this — sentiment classifiers don't capture the action space, and general intent classifiers don't understand the resize / reprice / defer distinctions specific to trade decisions.
|
| 23 |
+
|
| 24 |
+
This dataset fills that gap with a balanced, hand-curated 6-label corpus covering canonical replies and adversarial cases. The taxonomy and seeds were iterated against a working trading-agent reply distribution over multiple training cycles.
|
| 25 |
+
|
| 26 |
+
## TL;DR
|
| 27 |
+
|
| 28 |
+
- **25,200 training rows** generated by augmenting 1,047 human-written seed phrases via case / punctuation / whitespace and structural-prefix expansion (~24× per seed). Seeds are natural reply phrasings, not LLM-generated text.
|
| 29 |
+
- **175 held-out eval rows**, hand-curated, adversarial-leaning, zero-leakage verified against training.
|
| 30 |
+
- **6 balanced classes** — between 4,080 and 4,608 rows each.
|
| 31 |
+
- **Reproducible** — re-running the generator script with `--seed-rng 42` produces byte-identical output.
|
| 32 |
+
|
| 33 |
+
## Class distribution
|
| 34 |
+
|
| 35 |
+
| Label | Train | Eval | What it covers |
|
| 36 |
+
| -------------- | ------ | ---- | -------------- |
|
| 37 |
+
| APPROVE | 4,152 | 30 | Execute as proposed. "yes", "approve", "let's go" |
|
| 38 |
+
| DECLINE | 4,128 | 30 | Kill the proposal. "no", "pass", "kill it" |
|
| 39 |
+
| HOLD | 4,152 | 34 | Active deferral. "hold off", "checking", "leaning approve" |
|
| 40 |
+
| COUNTER_SIZE | 4,080 | 30 | Approve at different share count. "size 10", "dump half" |
|
| 41 |
+
| COUNTER_PRICE | 4,080 | 25 | Approve at different limit price. "at $49", "limit 50" |
|
| 42 |
+
| UNCLEAR | 4,608 | 26 | No committable position. Multi-intent, off-topic, ambiguous. |
|
| 43 |
+
|
| 44 |
+
UNCLEAR is slightly larger because it includes template-generated multi-intent examples using a 100-ticker pool (see Variation dimensions below).
|
| 45 |
+
|
| 46 |
+
## Schema
|
| 47 |
+
|
| 48 |
+
Each row is a JSON object:
|
| 49 |
+
|
| 50 |
+
```json
|
| 51 |
+
{
|
| 52 |
+
"text": "[dm][reply_to:131967][in_flight:1] Reject!",
|
| 53 |
+
"label": "DECLINE",
|
| 54 |
+
"label_id": 1,
|
| 55 |
+
"seed": "reject"
|
| 56 |
+
}
|
| 57 |
+
```
|
| 58 |
+
|
| 59 |
+
Fields:
|
| 60 |
+
|
| 61 |
+
- `text` — the prefixed input fed to the model
|
| 62 |
+
- `label` — string label
|
| 63 |
+
- `label_id` — int [0, 5]
|
| 64 |
+
- `seed` (train only) — original seed phrase, or `_ticker_tmpl_<...>` for template-generated rows
|
| 65 |
+
- `raw_text` (eval only) — original text before prefix wrap
|
| 66 |
+
- `note` (eval only) — curator's category note
|
| 67 |
+
|
| 68 |
+
Label ID mapping:
|
| 69 |
+
|
| 70 |
+
```
|
| 71 |
+
{"APPROVE": 0, "DECLINE": 1, "HOLD": 2,
|
| 72 |
+
"COUNTER_SIZE": 3, "COUNTER_PRICE": 4, "UNCLEAR": 5}
|
| 73 |
+
```
|
| 74 |
+
|
| 75 |
+
## Structural prefix scheme
|
| 76 |
+
|
| 77 |
+
Every row's `text` field starts with three tags carrying chat context:
|
| 78 |
+
|
| 79 |
+
```
|
| 80 |
+
[dm|group][reply_to:N|no_reply_to][in_flight:K]
|
| 81 |
+
```
|
| 82 |
+
|
| 83 |
+
These give the model context that pure text would not — "yes" with 1 proposal in flight is APPROVE; "yes" with 3 in flight and no quote-reply is structurally ambiguous and trained as UNCLEAR.
|
| 84 |
+
|
| 85 |
+
The prefix is concatenated into the input string and tokenized as regular subword pieces — no special-token registration required.
|
| 86 |
+
|
| 87 |
+
## Variation dimensions
|
| 88 |
+
|
| 89 |
+
Each of the 1,047 human-written seeds expands to ~24 training rows via two stacked variation axes:
|
| 90 |
+
|
| 91 |
+
**1. Surface variants (6 per seed, randomly sampled):**
|
| 92 |
+
|
| 93 |
+
- Case: lowercase / title / uppercase / as_is
|
| 94 |
+
- Punctuation: none / `.` / `!` / `?`
|
| 95 |
+
- Whitespace: none / leading 2 spaces / trailing 2 spaces
|
| 96 |
+
|
| 97 |
+
**2. Structural prefix variants (4 per surface form):**
|
| 98 |
+
|
| 99 |
+
- `(reply_to: present, in_flight: 1)`
|
| 100 |
+
- `(reply_to: present, in_flight: 2)`
|
| 101 |
+
- `(reply_to: present, in_flight: 3)`
|
| 102 |
+
- `(reply_to: absent, in_flight: 1)` — single-default rule
|
| 103 |
+
|
| 104 |
+
Additionally, the UNCLEAR class includes **480 template-generated rows** using a 100-ticker pool (mega-caps, ETFs, ADRs, mid-caps, speculative names) and 20 multi-intent templates ("approve {A} not {B}", "yes {A} no {B}", "swap {A} for {B}", etc.). This teaches the model the multi-intent PATTERN across diverse ticker tokens without bias toward any specific ticker. The model generalizes to ticker pairs it has never seen in training.
|
| 105 |
+
|
| 106 |
+
## Zero-leakage verification
|
| 107 |
+
|
| 108 |
+
Eval phrases are NEVER present in training:
|
| 109 |
+
|
| 110 |
+
```python
|
| 111 |
+
import json
|
| 112 |
+
|
| 113 |
+
training_seeds = set(
|
| 114 |
+
json.loads(l)["seed"].strip().lower()
|
| 115 |
+
for l in open("training.jsonl")
|
| 116 |
+
)
|
| 117 |
+
overlaps = sum(
|
| 118 |
+
1 for l in open("eval.jsonl")
|
| 119 |
+
if json.loads(l)["raw_text"].strip().lower() in training_seeds
|
| 120 |
+
)
|
| 121 |
+
assert overlaps == 0
|
| 122 |
+
```
|
| 123 |
+
|
| 124 |
+
This is enforced on every regeneration. The eval set's adversarial value depends on the model learning patterns it can generalize to novel phrasings, not memorizing surface forms.
|
| 125 |
+
|
| 126 |
+
## Design decisions
|
| 127 |
+
|
| 128 |
+
**Author-written seeds, not LLM-generated.** The 1,047 seed phrases reflect natural reply phrasings a human would actually type — short, casual, sometimes with typos, sometimes with adjacency to system commands. Seeds were hand-written and iteratively reviewed against eval failures.
|
| 129 |
+
|
| 130 |
+
**Programmatic augmentation, not LLM expansion.** Surface variation is deterministic — case/punctuation/whitespace flipping plus structural-prefix combinations. No language model generates content. This keeps the dataset auditable and reproducible.
|
| 131 |
+
|
| 132 |
+
**Six labels with COUNTER_PRICE separated from COUNTER_SIZE.** Earlier prototypes used five labels. The sixth (COUNTER_PRICE) was added because resize ("size 10") and reprice ("at $49") require different downstream extraction — share count vs limit price — and conflating them would force the consumer to disambiguate post-classification.
|
| 133 |
+
|
| 134 |
+
**Balanced classes by design.** All classes are within 4,080–4,608 rows (~13% spread). Class weighting in training becomes functionally uniform, simplifying the training loss and removing a configuration knob.
|
| 135 |
+
|
| 136 |
+
**Eval set adversarial-leaning.** The held-out 175 examples lean toward edge cases that surfaced during iterative training — casual register ("yup", "yeppers"), action metaphors ("press the button"), negation-as-deferral ("not now", "dont fire yet"), and multi-intent ambiguity ("yes but actually no"). Easy canonical examples are under-represented in eval intentionally.
|
| 137 |
+
|
| 138 |
+
## Intended downstream task
|
| 139 |
+
|
| 140 |
+
Train a classifier that routes short free-form text replies on trade proposals into one of 6 action intents:
|
| 141 |
+
|
| 142 |
+
```
|
| 143 |
+
APPROVE execute as proposed
|
| 144 |
+
DECLINE kill the proposal
|
| 145 |
+
HOLD defer the decision
|
| 146 |
+
COUNTER_SIZE execute at a different share count
|
| 147 |
+
COUNTER_PRICE execute at a different limit price
|
| 148 |
+
UNCLEAR cannot safely commit (refuse)
|
| 149 |
+
```
|
| 150 |
+
|
| 151 |
+
The data assumes the consumer pairs the model with a confidence threshold, deterministic safety rails (budget/position limits), and a fallback confirmation mechanism. The dataset is NOT intended for standalone-classifier deployments without those layers.
|
| 152 |
+
|
| 153 |
+
## Loading
|
| 154 |
+
|
| 155 |
+
```python
|
| 156 |
+
from datasets import load_dataset
|
| 157 |
+
ds = load_dataset("DoDataThings/trade-decision-classifier-v1-dataset")
|
| 158 |
+
ds["train"] # 25,200 rows
|
| 159 |
+
ds["test"] # 175 rows (held out, used for eval — never trained on)
|
| 160 |
+
```
|
| 161 |
+
|
| 162 |
+
Or directly:
|
| 163 |
+
|
| 164 |
+
```python
|
| 165 |
+
import json
|
| 166 |
+
train = [json.loads(l) for l in open("training.jsonl")]
|
| 167 |
+
test = [json.loads(l) for l in open("eval.jsonl")]
|
| 168 |
+
```
|
| 169 |
+
|
| 170 |
+
## Trained model
|
| 171 |
+
|
| 172 |
+
A reference classifier trained on this dataset: [DoDataThings/distilbert-trade-decision-classifier-v1](https://huggingface.co/DoDataThings/distilbert-trade-decision-classifier-v1). Reaches macro F1 0.954 on the held-out eval set with zero high-confidence misclassifications.
|
| 173 |
+
|
| 174 |
+
## Limitations
|
| 175 |
+
|
| 176 |
+
- **Surface coverage.** The 1,047 human-written seeds reflect representative reply phrasings curated through 12 training iterations against an adversarial eval set. Real production replies will still exceed seed coverage on long-tail phrasing; expect ~5% of replies to land at the confidence floor and route to confirmation fallback.
|
| 177 |
+
- **Class boundaries on multi-intent inputs** ("approve but cut size") are ambiguous; eval labels reflect a single curator's judgment.
|
| 178 |
+
- The "more shares" eval row is architecturally ambiguous (resize intent without a number) — a model trained on this dataset will correctly emit UNCLEAR rather than COUNTER_SIZE, which is safe behavior but reads as a miss against the eval label.
|
| 179 |
+
- **English-only.** No localization in v1.
|
| 180 |
+
|
| 181 |
+
## License
|
| 182 |
+
|
| 183 |
+
Apache 2.0.
|
eval.jsonl
ADDED
|
@@ -0,0 +1,175 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{"text": "[dm][reply_to:899950][in_flight:1] yeah send it", "label": "APPROVE", "label_id": 0, "raw_text": "yeah send it", "note": "casual affirm"}
|
| 2 |
+
{"text": "[dm][no_reply_to][in_flight:1] go go go", "label": "APPROVE", "label_id": 0, "raw_text": "go go go", "note": "triple-go single-default"}
|
| 3 |
+
{"text": "[dm][reply_to:899952][in_flight:1] do it now", "label": "APPROVE", "label_id": 0, "raw_text": "do it now", "note": "urgency variant"}
|
| 4 |
+
{"text": "[dm][reply_to:899953][in_flight:1] let's roll", "label": "APPROVE", "label_id": 0, "raw_text": "let's roll", "note": "idiom"}
|
| 5 |
+
{"text": "[dm][reply_to:899954][in_flight:1] I'm in", "label": "APPROVE", "label_id": 0, "raw_text": "I'm in", "note": "commit phrase"}
|
| 6 |
+
{"text": "[dm][reply_to:899955][in_flight:1] green light", "label": "APPROVE", "label_id": 0, "raw_text": "green light", "note": "metaphor"}
|
| 7 |
+
{"text": "[dm][reply_to:899956][in_flight:1] fire away", "label": "APPROVE", "label_id": 0, "raw_text": "fire away", "note": "metaphor variant"}
|
| 8 |
+
{"text": "[dm][reply_to:899957][in_flight:1] pull the trigger", "label": "APPROVE", "label_id": 0, "raw_text": "pull the trigger", "note": "metaphor"}
|
| 9 |
+
{"text": "[dm][reply_to:899958][in_flight:1] execute now", "label": "APPROVE", "label_id": 0, "raw_text": "execute now", "note": "explicit"}
|
| 10 |
+
{"text": "[dm][reply_to:899959][in_flight:1] press the button", "label": "APPROVE", "label_id": 0, "raw_text": "press the button", "note": "metaphor"}
|
| 11 |
+
{"text": "[dm][no_reply_to][in_flight:1] alright go", "label": "APPROVE", "label_id": 0, "raw_text": "alright go", "note": "mid-thought"}
|
| 12 |
+
{"text": "[dm][reply_to:899961][in_flight:1] yes I'm convinced", "label": "APPROVE", "label_id": 0, "raw_text": "yes I'm convinced", "note": "elaborated yes"}
|
| 13 |
+
{"text": "[dm][reply_to:899962][in_flight:1] we're good", "label": "APPROVE", "label_id": 0, "raw_text": "we're good", "note": "we're-form"}
|
| 14 |
+
{"text": "[dm][reply_to:899963][in_flight:1] send the order", "label": "APPROVE", "label_id": 0, "raw_text": "send the order", "note": "explicit action"}
|
| 15 |
+
{"text": "[dm][reply_to:899964][in_flight:1] approve approve", "label": "APPROVE", "label_id": 0, "raw_text": "approve approve", "note": "stutter"}
|
| 16 |
+
{"text": "[dm][reply_to:899965][in_flight:1] ya rly", "label": "APPROVE", "label_id": 0, "raw_text": "ya rly", "note": "casual emphatic"}
|
| 17 |
+
{"text": "[dm][reply_to:899966][in_flight:1] yeppers", "label": "APPROVE", "label_id": 0, "raw_text": "yeppers", "note": "playful yep"}
|
| 18 |
+
{"text": "[dm][reply_to:899967][in_flight:1] ya", "label": "APPROVE", "label_id": 0, "raw_text": "ya", "note": "short variant"}
|
| 19 |
+
{"text": "[dm][reply_to:899968][in_flight:1] yup", "label": "APPROVE", "label_id": 0, "raw_text": "yup", "note": "yup variant"}
|
| 20 |
+
{"text": "[dm][reply_to:899969][in_flight:1] aye", "label": "APPROVE", "label_id": 0, "raw_text": "aye", "note": "informal"}
|
| 21 |
+
{"text": "[dm][reply_to:899970][in_flight:1] affirmative", "label": "APPROVE", "label_id": 0, "raw_text": "affirmative", "note": "formal"}
|
| 22 |
+
{"text": "[dm][reply_to:899971][in_flight:1] grab it", "label": "APPROVE", "label_id": 0, "raw_text": "grab it", "note": "grab imperative"}
|
| 23 |
+
{"text": "[dm][reply_to:899972][in_flight:1] punch it", "label": "APPROVE", "label_id": 0, "raw_text": "punch it", "note": "slang"}
|
| 24 |
+
{"text": "[dm][reply_to:899973][in_flight:1] send it bro", "label": "APPROVE", "label_id": 0, "raw_text": "send it bro", "note": "casual address"}
|
| 25 |
+
{"text": "[dm][reply_to:899974][in_flight:1] buy", "label": "APPROVE", "label_id": 0, "raw_text": "buy", "note": "single-word action"}
|
| 26 |
+
{"text": "[dm][reply_to:899975][in_flight:1] approve and execute", "label": "APPROVE", "label_id": 0, "raw_text": "approve and execute", "note": "compound but same-intent"}
|
| 27 |
+
{"text": "[dm][reply_to:899976][in_flight:1] good to go", "label": "APPROVE", "label_id": 0, "raw_text": "good to go", "note": "idiom"}
|
| 28 |
+
{"text": "[dm][reply_to:899977][in_flight:1] all good execute", "label": "APPROVE", "label_id": 0, "raw_text": "all good execute", "note": "compound"}
|
| 29 |
+
{"text": "[dm][reply_to:899978][in_flight:1] yes lets do it", "label": "APPROVE", "label_id": 0, "raw_text": "yes lets do it", "note": "elaborated"}
|
| 30 |
+
{"text": "[dm][reply_to:899979][in_flight:1] lgtm send", "label": "APPROVE", "label_id": 0, "raw_text": "lgtm send", "note": "tech lingo"}
|
| 31 |
+
{"text": "[dm][reply_to:899980][in_flight:1] nah skip", "label": "DECLINE", "label_id": 1, "raw_text": "nah skip", "note": "casual + skip"}
|
| 32 |
+
{"text": "[dm][reply_to:899981][in_flight:1] no I don't like it", "label": "DECLINE", "label_id": 1, "raw_text": "no I don't like it", "note": "elaborated"}
|
| 33 |
+
{"text": "[dm][reply_to:899982][in_flight:1] pass on this", "label": "DECLINE", "label_id": 1, "raw_text": "pass on this", "note": "specific pass"}
|
| 34 |
+
{"text": "[dm][reply_to:899983][in_flight:1] actually nope", "label": "DECLINE", "label_id": 1, "raw_text": "actually nope", "note": "mind-change"}
|
| 35 |
+
{"text": "[dm][reply_to:899984][in_flight:1] no it's risky", "label": "DECLINE", "label_id": 1, "raw_text": "no it's risky", "note": "decline w/ reason same intent"}
|
| 36 |
+
{"text": "[dm][reply_to:899985][in_flight:1] chart isn't clean enough", "label": "DECLINE", "label_id": 1, "raw_text": "chart isn't clean enough", "note": "context decline"}
|
| 37 |
+
{"text": "[dm][reply_to:899986][in_flight:1] don't take it", "label": "DECLINE", "label_id": 1, "raw_text": "don't take it", "note": "imperative"}
|
| 38 |
+
{"text": "[dm][reply_to:899987][in_flight:1] scrap it", "label": "DECLINE", "label_id": 1, "raw_text": "scrap it", "note": "scrap variant"}
|
| 39 |
+
{"text": "[dm][reply_to:899988][in_flight:1] ditch it", "label": "DECLINE", "label_id": 1, "raw_text": "ditch it", "note": "ditch variant"}
|
| 40 |
+
{"text": "[dm][reply_to:899989][in_flight:1] nuke it", "label": "DECLINE", "label_id": 1, "raw_text": "nuke it", "note": "slang kill"}
|
| 41 |
+
{"text": "[dm][reply_to:899990][in_flight:1] kill", "label": "DECLINE", "label_id": 1, "raw_text": "kill", "note": "single-word kill"}
|
| 42 |
+
{"text": "[dm][reply_to:899991][in_flight:1] nope nope", "label": "DECLINE", "label_id": 1, "raw_text": "nope nope", "note": "double-nope"}
|
| 43 |
+
{"text": "[dm][reply_to:899992][in_flight:1] not interested", "label": "DECLINE", "label_id": 1, "raw_text": "not interested", "note": "specific"}
|
| 44 |
+
{"text": "[dm][reply_to:899993][in_flight:1] walk away", "label": "DECLINE", "label_id": 1, "raw_text": "walk away", "note": "metaphor"}
|
| 45 |
+
{"text": "[dm][reply_to:899994][in_flight:1] hard no", "label": "DECLINE", "label_id": 1, "raw_text": "hard no", "note": "hard prefix"}
|
| 46 |
+
{"text": "[dm][reply_to:899995][in_flight:1] absolutely not", "label": "DECLINE", "label_id": 1, "raw_text": "absolutely not", "note": "emphatic"}
|
| 47 |
+
{"text": "[dm][reply_to:899996][in_flight:1] no way", "label": "DECLINE", "label_id": 1, "raw_text": "no way", "note": "common"}
|
| 48 |
+
{"text": "[dm][reply_to:899997][in_flight:1] declining", "label": "DECLINE", "label_id": 1, "raw_text": "declining", "note": "verbed"}
|
| 49 |
+
{"text": "[dm][reply_to:899998][in_flight:1] rejected", "label": "DECLINE", "label_id": 1, "raw_text": "rejected", "note": "verbed"}
|
| 50 |
+
{"text": "[dm][reply_to:899999][in_flight:1] trash", "label": "DECLINE", "label_id": 1, "raw_text": "trash", "note": "single-word"}
|
| 51 |
+
{"text": "[dm][reply_to:900000][in_flight:1] garbage proposal", "label": "DECLINE", "label_id": 1, "raw_text": "garbage proposal", "note": "judgmental decline"}
|
| 52 |
+
{"text": "[dm][reply_to:900001][in_flight:1] no thx", "label": "DECLINE", "label_id": 1, "raw_text": "no thx", "note": "abbrev"}
|
| 53 |
+
{"text": "[dm][reply_to:900002][in_flight:1] nty", "label": "DECLINE", "label_id": 1, "raw_text": "nty", "note": "abbrev"}
|
| 54 |
+
{"text": "[dm][no_reply_to][in_flight:1] im out", "label": "DECLINE", "label_id": 1, "raw_text": "im out", "note": "im-form"}
|
| 55 |
+
{"text": "[dm][reply_to:900004][in_flight:1] opt out", "label": "DECLINE", "label_id": 1, "raw_text": "opt out", "note": "opt phrase"}
|
| 56 |
+
{"text": "[dm][reply_to:900005][in_flight:1] bail", "label": "DECLINE", "label_id": 1, "raw_text": "bail", "note": "single-word slang"}
|
| 57 |
+
{"text": "[dm][reply_to:900006][in_flight:1] dont buy", "label": "DECLINE", "label_id": 1, "raw_text": "dont buy", "note": "imperative neg"}
|
| 58 |
+
{"text": "[dm][reply_to:900007][in_flight:1] hard pass on this one", "label": "DECLINE", "label_id": 1, "raw_text": "hard pass on this one", "note": "elaborated pass"}
|
| 59 |
+
{"text": "[dm][reply_to:900008][in_flight:1] not feeling it", "label": "DECLINE", "label_id": 1, "raw_text": "not feeling it", "note": "feeling-form"}
|
| 60 |
+
{"text": "[dm][reply_to:900009][in_flight:1] sit this out", "label": "DECLINE", "label_id": 1, "raw_text": "sit this out", "note": "metaphor"}
|
| 61 |
+
{"text": "[dm][reply_to:900010][in_flight:1] let me look at the chart first", "label": "HOLD", "label_id": 2, "raw_text": "let me look at the chart first", "note": "explicit"}
|
| 62 |
+
{"text": "[dm][reply_to:900011][in_flight:1] still checking", "label": "HOLD", "label_id": 2, "raw_text": "still checking", "note": "verbed continuing"}
|
| 63 |
+
{"text": "[dm][reply_to:900012][in_flight:1] I need 5 minutes", "label": "HOLD", "label_id": 2, "raw_text": "I need 5 minutes", "note": "time-specific"}
|
| 64 |
+
{"text": "[dm][reply_to:900013][in_flight:1] hold til open", "label": "HOLD", "label_id": 2, "raw_text": "hold til open", "note": "conditional"}
|
| 65 |
+
{"text": "[dm][reply_to:900014][in_flight:1] wait until close", "label": "HOLD", "label_id": 2, "raw_text": "wait until close", "note": "conditional"}
|
| 66 |
+
{"text": "[dm][reply_to:900015][in_flight:1] wait for the pullback", "label": "HOLD", "label_id": 2, "raw_text": "wait for the pullback", "note": "conditional"}
|
| 67 |
+
{"text": "[dm][reply_to:900016][in_flight:1] after lunch", "label": "HOLD", "label_id": 2, "raw_text": "after lunch", "note": "time-deferred"}
|
| 68 |
+
{"text": "[dm][reply_to:900017][in_flight:1] end of day", "label": "HOLD", "label_id": 2, "raw_text": "end of day", "note": "time-deferred"}
|
| 69 |
+
{"text": "[dm][reply_to:900018][in_flight:1] come back to me", "label": "HOLD", "label_id": 2, "raw_text": "come back to me", "note": "deferral"}
|
| 70 |
+
{"text": "[dm][reply_to:900019][in_flight:1] ping me later", "label": "HOLD", "label_id": 2, "raw_text": "ping me later", "note": "deferral"}
|
| 71 |
+
{"text": "[dm][reply_to:900020][in_flight:1] standby", "label": "HOLD", "label_id": 2, "raw_text": "standby", "note": "imperative"}
|
| 72 |
+
{"text": "[dm][reply_to:900021][in_flight:1] stand by", "label": "HOLD", "label_id": 2, "raw_text": "stand by", "note": "imperative variant"}
|
| 73 |
+
{"text": "[dm][reply_to:900022][in_flight:1] queue it", "label": "HOLD", "label_id": 2, "raw_text": "queue it", "note": "queue metaphor"}
|
| 74 |
+
{"text": "[dm][reply_to:900023][in_flight:1] not now", "label": "HOLD", "label_id": 2, "raw_text": "not now", "note": "not-now"}
|
| 75 |
+
{"text": "[dm][reply_to:900024][in_flight:1] later today", "label": "HOLD", "label_id": 2, "raw_text": "later today", "note": "time-vague"}
|
| 76 |
+
{"text": "[dm][reply_to:900025][in_flight:1] hmm need to think", "label": "HOLD", "label_id": 2, "raw_text": "hmm need to think", "note": "thought-process"}
|
| 77 |
+
{"text": "[dm][reply_to:900026][in_flight:1] processing", "label": "HOLD", "label_id": 2, "raw_text": "processing", "note": "self-state"}
|
| 78 |
+
{"text": "[dm][reply_to:900027][in_flight:1] dont fire yet", "label": "HOLD", "label_id": 2, "raw_text": "dont fire yet", "note": "negative-imperative"}
|
| 79 |
+
{"text": "[dm][reply_to:900028][in_flight:1] not ready", "label": "HOLD", "label_id": 2, "raw_text": "not ready", "note": "readiness"}
|
| 80 |
+
{"text": "[dm][reply_to:900029][in_flight:1] checking levels", "label": "HOLD", "label_id": 2, "raw_text": "checking levels", "note": "specific action"}
|
| 81 |
+
{"text": "[dm][reply_to:900030][in_flight:1] reading the brief", "label": "HOLD", "label_id": 2, "raw_text": "reading the brief", "note": "specific action"}
|
| 82 |
+
{"text": "[dm][reply_to:900031][in_flight:1] in review", "label": "HOLD", "label_id": 2, "raw_text": "in review", "note": "passive variant"}
|
| 83 |
+
{"text": "[dm][reply_to:900032][in_flight:1] on it give me a min", "label": "HOLD", "label_id": 2, "raw_text": "on it give me a min", "note": "compound"}
|
| 84 |
+
{"text": "[dm][reply_to:900033][in_flight:1] hold pls", "label": "HOLD", "label_id": 2, "raw_text": "hold pls", "note": "abbrev pls"}
|
| 85 |
+
{"text": "[dm][reply_to:900034][in_flight:1] on hold", "label": "HOLD", "label_id": 2, "raw_text": "on hold", "note": "passive"}
|
| 86 |
+
{"text": "[dm][reply_to:900035][in_flight:1] pls wait", "label": "HOLD", "label_id": 2, "raw_text": "pls wait", "note": "abbrev wait"}
|
| 87 |
+
{"text": "[dm][reply_to:900036][in_flight:1] dont execute yet", "label": "HOLD", "label_id": 2, "raw_text": "dont execute yet", "note": "explicit defer"}
|
| 88 |
+
{"text": "[dm][reply_to:900037][in_flight:1] not yet pls", "label": "HOLD", "label_id": 2, "raw_text": "not yet pls", "note": "pls-suffixed"}
|
| 89 |
+
{"text": "[dm][reply_to:900038][in_flight:1] holding for now", "label": "HOLD", "label_id": 2, "raw_text": "holding for now", "note": "ing-form for-now"}
|
| 90 |
+
{"text": "[dm][reply_to:900039][in_flight:1] watching", "label": "HOLD", "label_id": 2, "raw_text": "watching", "note": "watching state"}
|
| 91 |
+
{"text": "[dm][reply_to:900040][in_flight:1] go with 7", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "go with 7", "note": "go-form"}
|
| 92 |
+
{"text": "[dm][reply_to:900041][in_flight:1] do twenty shares", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "do twenty shares", "note": "spelled-out"}
|
| 93 |
+
{"text": "[dm][reply_to:900042][in_flight:1] try ten", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "try ten", "note": "try-form spelled"}
|
| 94 |
+
{"text": "[dm][reply_to:900043][in_flight:1] only 5", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "only 5", "note": "only-form"}
|
| 95 |
+
{"text": "[dm][reply_to:900044][in_flight:1] just 3 sh", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "just 3 sh", "note": "just-form sh"}
|
| 96 |
+
{"text": "[dm][reply_to:900045][in_flight:1] bump it to 25", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "bump it to 25", "note": "bump-form"}
|
| 97 |
+
{"text": "[dm][reply_to:900046][in_flight:1] smaller make it 5", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "smaller make it 5", "note": "compound smaller"}
|
| 98 |
+
{"text": "[dm][reply_to:900047][in_flight:1] size 8 please", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "size 8 please", "note": "polite"}
|
| 99 |
+
{"text": "[dm][reply_to:900048][in_flight:1] can we do 6", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "can we do 6", "note": "question-form"}
|
| 100 |
+
{"text": "[dm][reply_to:900049][in_flight:1] how about 4 shares", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "how about 4 shares", "note": "how-about"}
|
| 101 |
+
{"text": "[dm][reply_to:900050][in_flight:1] try 9", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "try 9", "note": "try numeric"}
|
| 102 |
+
{"text": "[dm][reply_to:900051][in_flight:1] change to 12", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "change to 12", "note": "change-form"}
|
| 103 |
+
{"text": "[dm][reply_to:900052][in_flight:1] update to 14", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "update to 14", "note": "update-form"}
|
| 104 |
+
{"text": "[dm][reply_to:900053][in_flight:1] approve at 6", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "approve at 6", "note": "approve-at"}
|
| 105 |
+
{"text": "[dm][reply_to:900054][in_flight:1] yes 8 shares", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "yes 8 shares", "note": "yes + size"}
|
| 106 |
+
{"text": "[dm][reply_to:900055][in_flight:1] ok 11", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "ok 11", "note": "ok + num"}
|
| 107 |
+
{"text": "[dm][reply_to:900056][in_flight:1] go 7 sh", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "go 7 sh", "note": "go + sh"}
|
| 108 |
+
{"text": "[dm][reply_to:900057][in_flight:1] trim to 4", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "trim to 4", "note": "trim verb"}
|
| 109 |
+
{"text": "[dm][reply_to:900058][in_flight:1] size 6 not 10", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "size 6 not 10", "note": "explicit not-comparison"}
|
| 110 |
+
{"text": "[dm][reply_to:900059][in_flight:1] shave it to 4", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "shave it to 4", "note": "shave verb"}
|
| 111 |
+
{"text": "[dm][reply_to:900060][in_flight:1] more shares", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "more shares", "note": "more + sh"}
|
| 112 |
+
{"text": "[dm][reply_to:900061][in_flight:1] scale down to 5", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "scale down to 5", "note": "scale verb"}
|
| 113 |
+
{"text": "[dm][reply_to:900062][in_flight:1] scale up to 25", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "scale up to 25", "note": "scale verb up"}
|
| 114 |
+
{"text": "[dm][reply_to:900063][in_flight:1] knock it down to 8", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "knock it down to 8", "note": "knock-down"}
|
| 115 |
+
{"text": "[dm][reply_to:900064][in_flight:1] bump up to 20", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "bump up to 20", "note": "bump-up"}
|
| 116 |
+
{"text": "[dm][reply_to:900065][in_flight:1] 8 instead", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "8 instead", "note": "instead-suffix"}
|
| 117 |
+
{"text": "[dm][reply_to:900066][in_flight:1] size 3", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "size 3", "note": "tiny size"}
|
| 118 |
+
{"text": "[dm][reply_to:900067][in_flight:1] max 10", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "max 10", "note": "max-form"}
|
| 119 |
+
{"text": "[dm][reply_to:900068][in_flight:1] min 5", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "min 5", "note": "min-form"}
|
| 120 |
+
{"text": "[dm][reply_to:900069][in_flight:1] 100 sh too big do 50", "label": "COUNTER_SIZE", "label_id": 3, "raw_text": "100 sh too big do 50", "note": "compound-counter"}
|
| 121 |
+
{"text": "[dm][reply_to:900070][in_flight:1] at $49 instead", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "at $49 instead", "note": "$ explicit + 'instead'"}
|
| 122 |
+
{"text": "[dm][reply_to:900071][in_flight:1] trim it at $49", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "trim it at $49", "note": "verb-it + $price"}
|
| 123 |
+
{"text": "[dm][reply_to:900072][in_flight:1] exit at $50.25", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "exit at $50.25", "note": "decimal $price"}
|
| 124 |
+
{"text": "[dm][reply_to:900073][in_flight:1] sell at $48", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "sell at $48", "note": "verb + $price"}
|
| 125 |
+
{"text": "[dm][reply_to:900074][in_flight:1] buy at $42.50", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "buy at $42.50", "note": "decimal price"}
|
| 126 |
+
{"text": "[dm][reply_to:900075][in_flight:1] $48.75 limit", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "$48.75 limit", "note": "decimal + limit keyword"}
|
| 127 |
+
{"text": "[dm][reply_to:900076][in_flight:1] limit at 48", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "limit at 48", "note": "limit keyword no $"}
|
| 128 |
+
{"text": "[dm][reply_to:900077][in_flight:1] limit 49.5", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "limit 49.5", "note": "limit + decimal"}
|
| 129 |
+
{"text": "[dm][reply_to:900078][in_flight:1] price 49.5", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "price 49.5", "note": "price keyword + decimal"}
|
| 130 |
+
{"text": "[dm][reply_to:900079][in_flight:1] price at 49", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "price at 49", "note": "price + at"}
|
| 131 |
+
{"text": "[dm][reply_to:900080][in_flight:1] set price to 50", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "set price to 50", "note": "set price to"}
|
| 132 |
+
{"text": "[dm][reply_to:900081][in_flight:1] change the limit to 49", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "change the limit to 49", "note": "change limit"}
|
| 133 |
+
{"text": "[dm][reply_to:900082][in_flight:1] move limit higher to 50", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "move limit higher to 50", "note": "directional + limit"}
|
| 134 |
+
{"text": "[dm][reply_to:900083][in_flight:1] bump limit up to 51", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "bump limit up to 51", "note": "bump limit"}
|
| 135 |
+
{"text": "[dm][reply_to:900084][in_flight:1] drop the limit to 47", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "drop the limit to 47", "note": "drop limit"}
|
| 136 |
+
{"text": "[dm][reply_to:900085][in_flight:1] make it $48", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "make it $48", "note": "make it $"}
|
| 137 |
+
{"text": "[dm][reply_to:900086][in_flight:1] do it at $49.25", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "do it at $49.25", "note": "do it at $"}
|
| 138 |
+
{"text": "[dm][reply_to:900087][in_flight:1] run at $48", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "run at $48", "note": "run at"}
|
| 139 |
+
{"text": "[dm][reply_to:900088][in_flight:1] execute it at $50", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "execute it at $50", "note": "execute-it at"}
|
| 140 |
+
{"text": "[dm][reply_to:900089][in_flight:1] fire it at $49", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "fire it at $49", "note": "fire-it at"}
|
| 141 |
+
{"text": "[dm][reply_to:900090][in_flight:1] trim at 49 instead of 48", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "trim at 49 instead of 48", "note": "explicit comparison"}
|
| 142 |
+
{"text": "[dm][reply_to:900091][in_flight:1] sell at $48 not $50", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "sell at $48 not $50", "note": "not-comparison"}
|
| 143 |
+
{"text": "[dm][reply_to:900092][in_flight:1] @$50.25", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "@$50.25", "note": "at-sign + decimal"}
|
| 144 |
+
{"text": "[dm][reply_to:900093][in_flight:1] @ $49.50", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "@ $49.50", "note": "at + space + $"}
|
| 145 |
+
{"text": "[dm][reply_to:900094][in_flight:1] approve at $49", "label": "COUNTER_PRICE", "label_id": 4, "raw_text": "approve at $49", "note": "approve + at + $"}
|
| 146 |
+
{"text": "[dm][reply_to:900095][in_flight:1] approve but only half", "label": "UNCLEAR", "label_id": 5, "raw_text": "approve but only half", "note": "multi-intent approve+resize"}
|
| 147 |
+
{"text": "[dm][reply_to:900096][in_flight:1] approve and pause", "label": "UNCLEAR", "label_id": 5, "raw_text": "approve and pause", "note": "approve + structural"}
|
| 148 |
+
{"text": "[dm][reply_to:900097][in_flight:1] looking but probably no", "label": "UNCLEAR", "label_id": 5, "raw_text": "looking but probably no", "note": "soft lean"}
|
| 149 |
+
{"text": "[dm][reply_to:900098][in_flight:1] leaning towards yes", "label": "HOLD", "label_id": 2, "raw_text": "leaning towards yes", "note": "active opinion-forming"}
|
| 150 |
+
{"text": "[dm][reply_to:900099][in_flight:1] tempted but not sure", "label": "UNCLEAR", "label_id": 5, "raw_text": "tempted but not sure", "note": "tempted soft"}
|
| 151 |
+
{"text": "[dm][reply_to:900100][in_flight:1] not sure yet", "label": "UNCLEAR", "label_id": 5, "raw_text": "not sure yet", "note": "not-sure"}
|
| 152 |
+
{"text": "[dm][reply_to:900101][in_flight:1] ???", "label": "UNCLEAR", "label_id": 5, "raw_text": "???", "note": "triple-question"}
|
| 153 |
+
{"text": "[dm][reply_to:900102][in_flight:1] what's the entry?", "label": "UNCLEAR", "label_id": 5, "raw_text": "what's the entry?", "note": "question-back"}
|
| 154 |
+
{"text": "[dm][reply_to:900103][in_flight:1] what's the stop", "label": "UNCLEAR", "label_id": 5, "raw_text": "what's the stop", "note": "question-back"}
|
| 155 |
+
{"text": "[dm][reply_to:900104][in_flight:1] why this size", "label": "UNCLEAR", "label_id": 5, "raw_text": "why this size", "note": "question-back"}
|
| 156 |
+
{"text": "[dm][reply_to:900105][in_flight:1] send the chart", "label": "UNCLEAR", "label_id": 5, "raw_text": "send the chart", "note": "off-topic request"}
|
| 157 |
+
{"text": "[dm][reply_to:900106][in_flight:1] dashboard", "label": "UNCLEAR", "label_id": 5, "raw_text": "dashboard", "note": "single-word off-topic"}
|
| 158 |
+
{"text": "[dm][reply_to:900107][in_flight:1] any news on this", "label": "UNCLEAR", "label_id": 5, "raw_text": "any news on this", "note": "general inquiry"}
|
| 159 |
+
{"text": "[dm][reply_to:900108][in_flight:1] how's the portfolio", "label": "UNCLEAR", "label_id": 5, "raw_text": "how's the portfolio", "note": "off-topic portfolio"}
|
| 160 |
+
{"text": "[dm][reply_to:900109][in_flight:1] k", "label": "UNCLEAR", "label_id": 5, "raw_text": "k", "note": "single-letter ambiguous"}
|
| 161 |
+
{"text": "[dm][reply_to:900110][in_flight:1] kk", "label": "UNCLEAR", "label_id": 5, "raw_text": "kk", "note": "kk ambiguous"}
|
| 162 |
+
{"text": "[dm][reply_to:900111][in_flight:1] ...", "label": "UNCLEAR", "label_id": 5, "raw_text": "...", "note": "ellipsis"}
|
| 163 |
+
{"text": "[dm][reply_to:900112][in_flight:1] yes but actually no", "label": "UNCLEAR", "label_id": 5, "raw_text": "yes but actually no", "note": "self-contradiction"}
|
| 164 |
+
{"text": "[dm][reply_to:900113][in_flight:1] no wait yes", "label": "UNCLEAR", "label_id": 5, "raw_text": "no wait yes", "note": "mind-change-mid"}
|
| 165 |
+
{"text": "[dm][reply_to:900114][in_flight:1] ill let you know", "label": "HOLD", "label_id": 2, "raw_text": "ill let you know", "note": "future-tense engagement"}
|
| 166 |
+
{"text": "[dm][reply_to:900115][in_flight:1] send it but smaller", "label": "UNCLEAR", "label_id": 5, "raw_text": "send it but smaller", "note": "approve+counter compound"}
|
| 167 |
+
{"text": "[dm][reply_to:900116][in_flight:1] approve the other one", "label": "UNCLEAR", "label_id": 5, "raw_text": "approve the other one", "note": "ambiguous referent"}
|
| 168 |
+
{"text": "[dm][reply_to:900117][in_flight:1] eh whatever", "label": "UNCLEAR", "label_id": 5, "raw_text": "eh whatever", "note": "dismissive ambiguous"}
|
| 169 |
+
{"text": "[dm][reply_to:900118][in_flight:1] idk you call it", "label": "UNCLEAR", "label_id": 5, "raw_text": "idk you call it", "note": "deferral to agent"}
|
| 170 |
+
{"text": "[dm][reply_to:900119][in_flight:1] typing", "label": "HOLD", "label_id": 2, "raw_text": "typing", "note": "active engagement state"}
|
| 171 |
+
{"text": "[dm][reply_to:900120][in_flight:1] brb thinking", "label": "HOLD", "label_id": 2, "raw_text": "brb thinking", "note": "active deferral"}
|
| 172 |
+
{"text": "[dm][reply_to:900121][in_flight:1] uh", "label": "UNCLEAR", "label_id": 5, "raw_text": "uh", "note": "filler"}
|
| 173 |
+
{"text": "[dm][reply_to:900122][in_flight:1] interesting", "label": "UNCLEAR", "label_id": 5, "raw_text": "interesting", "note": "evaluative not decision"}
|
| 174 |
+
{"text": "[dm][reply_to:900123][in_flight:1] good setup", "label": "UNCLEAR", "label_id": 5, "raw_text": "good setup", "note": "compliment not decision"}
|
| 175 |
+
{"text": "[dm][reply_to:900124][in_flight:1] looks decent", "label": "UNCLEAR", "label_id": 5, "raw_text": "looks decent", "note": "evaluative"}
|
seeds.yaml
ADDED
|
@@ -0,0 +1,1194 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Reply classifier seed vocabulary — Phase 2a
|
| 2 |
+
#
|
| 3 |
+
# These are the canonical surface phrases the model is trained on. The
|
| 4 |
+
# synthetic-data generator (scripts/classifier/gen_synth_data.py) applies
|
| 5 |
+
# programmatic expansion: case variants, punctuation, leading/trailing
|
| 6 |
+
# whitespace, and structural-prefix combinations. ~30 seeds × ~30
|
| 7 |
+
# expansions ≈ ~900 examples per label before LLM-assisted breadth
|
| 8 |
+
# augments the UNCLEAR class.
|
| 9 |
+
#
|
| 10 |
+
# The eval set (eval.jsonl) is held out separately — none of these
|
| 11 |
+
# phrases (or trivial variants thereof) should appear in eval. Winston
|
| 12 |
+
# reviews this file. Anything added here gets multiplied; treat each
|
| 13 |
+
# entry as a representative cluster, not a single phrase.
|
| 14 |
+
#
|
| 15 |
+
# Format:
|
| 16 |
+
# <LABEL>:
|
| 17 |
+
# core: # The canonical phrases. The model must classify these.
|
| 18 |
+
# - phrase
|
| 19 |
+
# adversarial: # Borderline phrases that ARE this label despite
|
| 20 |
+
# # looking like a neighbor (e.g. "yes go" looks like
|
| 21 |
+
# # "yes" + something).
|
| 22 |
+
# - phrase
|
| 23 |
+
#
|
| 24 |
+
# CROSSCLASS_NEGATIVES is the explicit UNCLEAR class for phrases that
|
| 25 |
+
# LOOK like APPROVE/DECLINE/HOLD/COUNTER_SIZE but aren't safe to commit.
|
| 26 |
+
# These are the hard examples — the model is trained to refuse here.
|
| 27 |
+
|
| 28 |
+
APPROVE:
|
| 29 |
+
core:
|
| 30 |
+
# Canonical affirmations
|
| 31 |
+
- approve
|
| 32 |
+
- "yes"
|
| 33 |
+
- ok
|
| 34 |
+
- okay
|
| 35 |
+
- execute
|
| 36 |
+
- go
|
| 37 |
+
- do it
|
| 38 |
+
- send it
|
| 39 |
+
- confirmed
|
| 40 |
+
- confirm
|
| 41 |
+
- proceed
|
| 42 |
+
- run it
|
| 43 |
+
- take it
|
| 44 |
+
- lets go
|
| 45 |
+
- lgtm
|
| 46 |
+
- approved
|
| 47 |
+
- yep
|
| 48 |
+
- sure
|
| 49 |
+
- 'y'
|
| 50 |
+
- go ahead
|
| 51 |
+
- fire it
|
| 52 |
+
- ship it
|
| 53 |
+
# Casual register — similar shape to eval but NOT eval rows
|
| 54 |
+
- yuppy
|
| 55 |
+
- yepyep
|
| 56 |
+
- yes for real
|
| 57 |
+
- yea
|
| 58 |
+
- aye aye
|
| 59 |
+
- we are go
|
| 60 |
+
- ready to roll
|
| 61 |
+
- count me in
|
| 62 |
+
- im on board
|
| 63 |
+
- im sold
|
| 64 |
+
- love it
|
| 65 |
+
- this is it
|
| 66 |
+
- buy buy buy
|
| 67 |
+
- lets ride
|
| 68 |
+
- lets rip
|
| 69 |
+
- let's send it
|
| 70 |
+
- lets do it
|
| 71 |
+
- I like it
|
| 72 |
+
- looks good
|
| 73 |
+
- looks great
|
| 74 |
+
- looks clean
|
| 75 |
+
# Action-metaphor approves — "fire / press / hit / trigger / punch"
|
| 76 |
+
- hit the button
|
| 77 |
+
- trigger it
|
| 78 |
+
- fire the order
|
| 79 |
+
- send the trade
|
| 80 |
+
- punch the order
|
| 81 |
+
- kick it off
|
| 82 |
+
- launch it
|
| 83 |
+
- pop it
|
| 84 |
+
- smash it
|
| 85 |
+
# "let's X" cluster — anti-bleed for HOLD's "let's defer"/"let's hold"
|
| 86 |
+
- let's send
|
| 87 |
+
- let's send now
|
| 88 |
+
- let's pull it
|
| 89 |
+
- let's execute
|
| 90 |
+
- let's run it
|
| 91 |
+
- let's take it
|
| 92 |
+
- let's punch it
|
| 93 |
+
- lets fire
|
| 94 |
+
- lets go now
|
| 95 |
+
- lets do this
|
| 96 |
+
# Formal / metaphor approves
|
| 97 |
+
- agreed
|
| 98 |
+
- concur
|
| 99 |
+
- confirm yes
|
| 100 |
+
- go signal
|
| 101 |
+
- clear to fire
|
| 102 |
+
- thumbs up
|
| 103 |
+
- acknowledged
|
| 104 |
+
# More casual abbrev variants — bridging to "yeppers"/"ya rly"/"yup" eval
|
| 105 |
+
- yepp
|
| 106 |
+
- ye
|
| 107 |
+
- yeahyeah
|
| 108 |
+
- yarly
|
| 109 |
+
- ya for sure
|
| 110 |
+
- sure thing
|
| 111 |
+
- bet
|
| 112 |
+
- fasho
|
| 113 |
+
- heck yes
|
| 114 |
+
# Action-imperative variants — bridging to "fire away"
|
| 115 |
+
- fire it off
|
| 116 |
+
- shoot it
|
| 117 |
+
- blast it
|
| 118 |
+
- go now
|
| 119 |
+
# MORE casual-register expansion — covers "yup"/"yeppers"/"good to go"
|
| 120 |
+
- yupper
|
| 121 |
+
- yups
|
| 122 |
+
- yup yup
|
| 123 |
+
- sup yup
|
| 124 |
+
- yep that
|
| 125 |
+
- yep send
|
| 126 |
+
- ya all the way
|
| 127 |
+
- ya for it
|
| 128 |
+
- yarr
|
| 129 |
+
- good to send
|
| 130 |
+
- good to fire
|
| 131 |
+
- clean to go
|
| 132 |
+
- clear to go
|
| 133 |
+
- ready set
|
| 134 |
+
- we are set
|
| 135 |
+
# MORE action-metaphor bridges — "press the button"/"pull the trigger"
|
| 136 |
+
- press it
|
| 137 |
+
- press it now
|
| 138 |
+
- push the button
|
| 139 |
+
- pull it
|
| 140 |
+
- pull trigger
|
| 141 |
+
- mash the button
|
| 142 |
+
- hit it
|
| 143 |
+
- hit go
|
| 144 |
+
- hit send
|
| 145 |
+
# MORE "lgtm" / dev-lingo
|
| 146 |
+
- lgtm
|
| 147 |
+
- lgtm fire
|
| 148 |
+
- lgtm execute
|
| 149 |
+
- looks good fire
|
| 150 |
+
- looks good go
|
| 151 |
+
- clean lgtm
|
| 152 |
+
- solid lgtm
|
| 153 |
+
- lgtm fire it
|
| 154 |
+
# "I'm in" cluster — anti-bleed from HOLD's "count me in"
|
| 155 |
+
- in for it
|
| 156 |
+
- im in for it
|
| 157 |
+
- yes im in
|
| 158 |
+
- ok im in
|
| 159 |
+
- imma do it
|
| 160 |
+
- imma execute
|
| 161 |
+
- imma send
|
| 162 |
+
- imma punch it
|
| 163 |
+
- sign me up
|
| 164 |
+
- im there
|
| 165 |
+
- im in on it
|
| 166 |
+
- im on this
|
| 167 |
+
- i am in
|
| 168 |
+
- im in dude
|
| 169 |
+
- fully in
|
| 170 |
+
- all the way in
|
| 171 |
+
# "green light" cluster — APPROVE on go-signal metaphors
|
| 172 |
+
- green it
|
| 173 |
+
- lit
|
| 174 |
+
- light it up
|
| 175 |
+
- give the green
|
| 176 |
+
- greenlight it
|
| 177 |
+
- ok green light
|
| 178 |
+
- light em up
|
| 179 |
+
# "rly" / "for real" emphasizers
|
| 180 |
+
- for real yes
|
| 181 |
+
- rly
|
| 182 |
+
- really yes
|
| 183 |
+
- yes really
|
| 184 |
+
- for real do it
|
| 185 |
+
- no really yes
|
| 186 |
+
- 100% yes
|
| 187 |
+
- hundred percent
|
| 188 |
+
# "all in" / "all out" / "full" — execute proposal as-proposed (NOT counter-size)
|
| 189 |
+
# On an exit proposal: "dump it all" = APPROVE (full exit, the default).
|
| 190 |
+
# On an entry proposal: "all in" = APPROVE (full entry at proposed size).
|
| 191 |
+
- all in
|
| 192 |
+
- all out
|
| 193 |
+
- dump it all
|
| 194 |
+
- dump everything
|
| 195 |
+
- dump all
|
| 196 |
+
- exit all
|
| 197 |
+
- exit everything
|
| 198 |
+
- close it out
|
| 199 |
+
- close the whole position
|
| 200 |
+
- send it all
|
| 201 |
+
- the full thing
|
| 202 |
+
- full position
|
| 203 |
+
- yes full
|
| 204 |
+
- yes 100%
|
| 205 |
+
- execute fully
|
| 206 |
+
- execute it all
|
| 207 |
+
- do the whole thing
|
| 208 |
+
adversarial:
|
| 209 |
+
- yes go
|
| 210 |
+
- yes please
|
| 211 |
+
- approve it
|
| 212 |
+
- ok proceed
|
| 213 |
+
- yeah do it
|
| 214 |
+
- alright execute
|
| 215 |
+
- approve this one
|
| 216 |
+
- approve this
|
| 217 |
+
- send this one
|
| 218 |
+
- take this one
|
| 219 |
+
- lets take it
|
| 220 |
+
|
| 221 |
+
DECLINE:
|
| 222 |
+
core:
|
| 223 |
+
# Canonical negatives
|
| 224 |
+
- decline
|
| 225 |
+
- "no"
|
| 226 |
+
- reject
|
| 227 |
+
- pass
|
| 228 |
+
- nope
|
| 229 |
+
- nah
|
| 230 |
+
- no thanks
|
| 231 |
+
- kill it
|
| 232 |
+
- skip it
|
| 233 |
+
- cancel it
|
| 234 |
+
- drop it
|
| 235 |
+
- decline it
|
| 236 |
+
- 'n'
|
| 237 |
+
- hard pass
|
| 238 |
+
- dont take it
|
| 239 |
+
- reject this
|
| 240 |
+
- not taking it
|
| 241 |
+
- not this one
|
| 242 |
+
- no thanks not this one
|
| 243 |
+
- kill the proposal
|
| 244 |
+
- cancel proposal
|
| 245 |
+
- skip
|
| 246 |
+
# Casual / colloquial — similar shape to eval but NOT eval rows
|
| 247 |
+
- bail out
|
| 248 |
+
- i am out
|
| 249 |
+
- im out of this
|
| 250 |
+
- imma pass
|
| 251 |
+
- imma bail
|
| 252 |
+
- skipping this one
|
| 253 |
+
- not my style
|
| 254 |
+
- no interest
|
| 255 |
+
- walking away
|
| 256 |
+
- no chance
|
| 257 |
+
- hard pass for me
|
| 258 |
+
- absolutely no
|
| 259 |
+
- n t y
|
| 260 |
+
- no thnx
|
| 261 |
+
- hard pass here
|
| 262 |
+
- thumbs down
|
| 263 |
+
- vetoed
|
| 264 |
+
# Anti-bleed on "this" / "on this" — keep these explicitly DECLINE
|
| 265 |
+
# so the model doesn't learn "this" + verb leans HOLD
|
| 266 |
+
- passing on this one
|
| 267 |
+
- going to pass on this
|
| 268 |
+
- reject this one
|
| 269 |
+
- decline this one
|
| 270 |
+
- drop this one
|
| 271 |
+
- kill this one
|
| 272 |
+
- this one isnt for me
|
| 273 |
+
- this trade not for me
|
| 274 |
+
- dropping this trade
|
| 275 |
+
- abandoning this one
|
| 276 |
+
- this isnt worth it
|
| 277 |
+
- fail this trade
|
| 278 |
+
- kill this trade
|
| 279 |
+
- drop this trade
|
| 280 |
+
# "not interested" cluster — anti-UNCLEAR bleed
|
| 281 |
+
- no interest in this
|
| 282 |
+
- uninterested
|
| 283 |
+
- not into it
|
| 284 |
+
- dont care for it
|
| 285 |
+
- not feeling this trade
|
| 286 |
+
# "im out" stubborn — needs more variation in shape
|
| 287 |
+
- i out
|
| 288 |
+
- ima out
|
| 289 |
+
- im out of this trade
|
| 290 |
+
- outta this
|
| 291 |
+
- outta this trade
|
| 292 |
+
- im out yall
|
| 293 |
+
- ima dip
|
| 294 |
+
# "nty" / abbreviation variants
|
| 295 |
+
- n thx
|
| 296 |
+
- no tnks
|
| 297 |
+
- notnks
|
| 298 |
+
- ntk
|
| 299 |
+
- n thanks
|
| 300 |
+
# "not interested" / interest-negation — stubborn DECLINE 3 runs running
|
| 301 |
+
- not into this
|
| 302 |
+
- not my kind of trade
|
| 303 |
+
- dont find this interesting
|
| 304 |
+
- no appeal
|
| 305 |
+
- not worth my attention
|
| 306 |
+
- meh decline
|
| 307 |
+
- meh skip
|
| 308 |
+
- dull pass
|
| 309 |
+
# "im out" stubborn — variations the model hasn't seen
|
| 310 |
+
- im so out
|
| 311 |
+
- im totally out
|
| 312 |
+
- im definitely out
|
| 313 |
+
- peace im out
|
| 314 |
+
- im done with this
|
| 315 |
+
- im done here
|
| 316 |
+
- opting out
|
| 317 |
+
- tapping out
|
| 318 |
+
# ─── Balancing additions (msg 908) — DECLINE variety ─────
|
| 319 |
+
- naw
|
| 320 |
+
- naw fam
|
| 321 |
+
- no go
|
| 322 |
+
- no go on this
|
| 323 |
+
- not going
|
| 324 |
+
- not going to do it
|
| 325 |
+
- wont do it
|
| 326 |
+
- wont take this
|
| 327 |
+
- abstain
|
| 328 |
+
- abstaining
|
| 329 |
+
- decline this trade
|
| 330 |
+
- passing on this trade
|
| 331 |
+
- stepping back
|
| 332 |
+
- stepping away
|
| 333 |
+
- i pass
|
| 334 |
+
- i decline
|
| 335 |
+
- i reject
|
| 336 |
+
- i reject this
|
| 337 |
+
- veto
|
| 338 |
+
- veto this
|
| 339 |
+
- veto this trade
|
| 340 |
+
- booo
|
| 341 |
+
- boooo
|
| 342 |
+
- nah dont
|
| 343 |
+
- nah dont take it
|
| 344 |
+
- no thanks pass
|
| 345 |
+
- decline thanks
|
| 346 |
+
- no for me
|
| 347 |
+
- no for me thanks
|
| 348 |
+
- not buying it
|
| 349 |
+
- not selling on this
|
| 350 |
+
- this is a no
|
| 351 |
+
- this is a no go
|
| 352 |
+
- no this one
|
| 353 |
+
- decline kill
|
| 354 |
+
- drop it kill
|
| 355 |
+
- dumpster fire pass
|
| 356 |
+
- awful pass
|
| 357 |
+
- terrible pass
|
| 358 |
+
adversarial:
|
| 359 |
+
- actually no
|
| 360 |
+
- no decline
|
| 361 |
+
- nah pass
|
| 362 |
+
- ill pass
|
| 363 |
+
- not for me thanks
|
| 364 |
+
- no thanks im out
|
| 365 |
+
- im not feeling it
|
| 366 |
+
- this isn't it
|
| 367 |
+
- doesn't look right
|
| 368 |
+
- chart looks bad
|
| 369 |
+
# Stronger "this" anti-bleed — "X this out" / "X this trade" patterns
|
| 370 |
+
- im sitting this trade out
|
| 371 |
+
- sitting out this one
|
| 372 |
+
- leaving this trade
|
| 373 |
+
- leaving this one
|
| 374 |
+
- letting this go
|
| 375 |
+
- letting this one go
|
| 376 |
+
- throwing this trade away
|
| 377 |
+
- not pursuing this
|
| 378 |
+
- not taking this trade
|
| 379 |
+
- im not taking this one
|
| 380 |
+
- skipping this trade
|
| 381 |
+
# "bail" cluster — anti-confusion with APPROVE
|
| 382 |
+
- bailing
|
| 383 |
+
- bailing on this
|
| 384 |
+
- im bailing
|
| 385 |
+
- gonna bail
|
| 386 |
+
- bail on this
|
| 387 |
+
- bail this trade
|
| 388 |
+
# Single-word destruction verbs — "nuke it" cluster
|
| 389 |
+
- trash it
|
| 390 |
+
- torch it
|
| 391 |
+
- shred it
|
| 392 |
+
- vaporize it
|
| 393 |
+
- dump it
|
| 394 |
+
- chuck it
|
| 395 |
+
- toss it
|
| 396 |
+
- junk it
|
| 397 |
+
- bin it
|
| 398 |
+
- flush it
|
| 399 |
+
# Stubborn "sit X out" / "this trade out" reinforcement
|
| 400 |
+
- sitting out
|
| 401 |
+
- sittin out
|
| 402 |
+
- sit it out
|
| 403 |
+
- going to sit
|
| 404 |
+
- im sitting
|
| 405 |
+
- sitting on the sidelines
|
| 406 |
+
- benching this
|
| 407 |
+
- benching this one
|
| 408 |
+
- shelving this
|
| 409 |
+
- shelving this trade
|
| 410 |
+
|
| 411 |
+
HOLD:
|
| 412 |
+
core:
|
| 413 |
+
# Explicit deferral
|
| 414 |
+
- hold
|
| 415 |
+
- hold it
|
| 416 |
+
- hold this
|
| 417 |
+
- hold this one
|
| 418 |
+
- hold this for now
|
| 419 |
+
- hold on
|
| 420 |
+
- hold off
|
| 421 |
+
- let's hold
|
| 422 |
+
- let's defer
|
| 423 |
+
- defer
|
| 424 |
+
- defer this
|
| 425 |
+
- defer for now
|
| 426 |
+
- wait
|
| 427 |
+
- wait on this
|
| 428 |
+
- not yet
|
| 429 |
+
- not yet decide
|
| 430 |
+
# Active state-of-attention (user IS engaged, just not deciding)
|
| 431 |
+
- let me think
|
| 432 |
+
- let me look
|
| 433 |
+
- thinking
|
| 434 |
+
- lemme think
|
| 435 |
+
- need to check
|
| 436 |
+
- need a minute
|
| 437 |
+
- need to look at chart
|
| 438 |
+
- bit more time
|
| 439 |
+
- give me a sec
|
| 440 |
+
- one moment
|
| 441 |
+
- looking
|
| 442 |
+
- looking at it
|
| 443 |
+
- reading
|
| 444 |
+
- reading the thesis
|
| 445 |
+
- reviewing
|
| 446 |
+
- under review
|
| 447 |
+
- checking
|
| 448 |
+
- watching it
|
| 449 |
+
- monitoring
|
| 450 |
+
- tracking it
|
| 451 |
+
- on my radar
|
| 452 |
+
- parking this
|
| 453 |
+
- park it
|
| 454 |
+
- leave it open
|
| 455 |
+
- sit on it
|
| 456 |
+
- thinking still
|
| 457 |
+
- thinking it over
|
| 458 |
+
- still mulling
|
| 459 |
+
- let me digest
|
| 460 |
+
# Future-tense engagement (user committing to return)
|
| 461 |
+
- ill get back to you
|
| 462 |
+
- circling back later
|
| 463 |
+
- ok let me see
|
| 464 |
+
- one sec
|
| 465 |
+
- brb
|
| 466 |
+
- on phone
|
| 467 |
+
- in a meeting
|
| 468 |
+
- back in 5
|
| 469 |
+
- cant talk now
|
| 470 |
+
- later
|
| 471 |
+
- tonight
|
| 472 |
+
- lemme dig in
|
| 473 |
+
- need a chart
|
| 474 |
+
- pull it up
|
| 475 |
+
- resend
|
| 476 |
+
# Active opinion-forming (user IS taking a "leaning" position)
|
| 477 |
+
- I think
|
| 478 |
+
- maybe
|
| 479 |
+
- leaning yes
|
| 480 |
+
- leaning approve
|
| 481 |
+
- leaning that way
|
| 482 |
+
- probably approve
|
| 483 |
+
- inclined to take it
|
| 484 |
+
- lean decline
|
| 485 |
+
- 50/50
|
| 486 |
+
adversarial:
|
| 487 |
+
# Adversarial: APPROVE-shaped tokens that flip on the HOLD modifier
|
| 488 |
+
- yes on hold
|
| 489 |
+
- go hold
|
| 490 |
+
- alright hold
|
| 491 |
+
- ok hold off
|
| 492 |
+
- yes hold this
|
| 493 |
+
# Negation-as-deferral: "dont X yet" is HOLD, not DECLINE
|
| 494 |
+
- dont buy yet
|
| 495 |
+
- dont act yet
|
| 496 |
+
- dont fire it yet
|
| 497 |
+
- dont send it yet
|
| 498 |
+
- not ready to commit
|
| 499 |
+
- wait pls
|
| 500 |
+
- please wait
|
| 501 |
+
- not now please
|
| 502 |
+
- hold please
|
| 503 |
+
- just wait
|
| 504 |
+
- lets wait
|
| 505 |
+
- hold this for a bit
|
| 506 |
+
- give it a moment
|
| 507 |
+
# "not X yet/now" cluster — HOLD not DECLINE (active deferral with negation)
|
| 508 |
+
- not yet now
|
| 509 |
+
- not for now
|
| 510 |
+
- not at the moment
|
| 511 |
+
- not this moment
|
| 512 |
+
- not ready right now
|
| 513 |
+
- not now actually wait
|
| 514 |
+
- this isnt the time
|
| 515 |
+
# "queue X" cluster — HOLD anti-bleed from DECLINE
|
| 516 |
+
- queue this
|
| 517 |
+
- queue this one
|
| 518 |
+
- queue it up
|
| 519 |
+
- queue for later
|
| 520 |
+
- queue this for later
|
| 521 |
+
- into the queue
|
| 522 |
+
- put in the queue
|
| 523 |
+
- add to queue
|
| 524 |
+
# "checking levels" / level/chart-check HOLD anti-bleed
|
| 525 |
+
- checking the level
|
| 526 |
+
- checking key levels
|
| 527 |
+
- looking at levels
|
| 528 |
+
- level check
|
| 529 |
+
- checking the chart
|
| 530 |
+
- checking key prices
|
| 531 |
+
- reviewing levels
|
| 532 |
+
# "ill let you know" / future-update HOLD
|
| 533 |
+
- ill text back
|
| 534 |
+
- ill update soon
|
| 535 |
+
- give me time ill respond
|
| 536 |
+
- ill ping you
|
| 537 |
+
- ill follow up
|
| 538 |
+
# "standby" / "stand by" — APPROVE-confusable HOLD
|
| 539 |
+
# Run #9 had these classified as APPROVE @ 0.36/0.48
|
| 540 |
+
- standby for now
|
| 541 |
+
- stand by for now
|
| 542 |
+
- standing by
|
| 543 |
+
- hold standby
|
| 544 |
+
- just standby
|
| 545 |
+
- stay
|
| 546 |
+
- stay tuned
|
| 547 |
+
- wait standby
|
| 548 |
+
- on standby
|
| 549 |
+
# ─── Balancing additions (msg 908) — HOLD variety ─────
|
| 550 |
+
- hold a moment
|
| 551 |
+
- hold a minute
|
| 552 |
+
- hold a sec
|
| 553 |
+
- give it a sec
|
| 554 |
+
- give it a min
|
| 555 |
+
- give it a moment
|
| 556 |
+
- give me a moment
|
| 557 |
+
- give me one min
|
| 558 |
+
- give me two min
|
| 559 |
+
- give me five min
|
| 560 |
+
- back soon
|
| 561 |
+
- back in a few
|
| 562 |
+
- be right back
|
| 563 |
+
- moment please
|
| 564 |
+
- second please
|
| 565 |
+
- lemme see
|
| 566 |
+
- lemme look
|
| 567 |
+
- let me see
|
| 568 |
+
- let me look it up
|
| 569 |
+
- look at it later
|
| 570 |
+
- look later
|
| 571 |
+
- checking now
|
| 572 |
+
- verifying
|
| 573 |
+
- verifying now
|
| 574 |
+
- thinking about it
|
| 575 |
+
- thinking on it
|
| 576 |
+
- thinking through
|
| 577 |
+
- pondering
|
| 578 |
+
- pondering it
|
| 579 |
+
- weighing
|
| 580 |
+
- weighing options
|
| 581 |
+
- mulling it over
|
| 582 |
+
- mulling
|
| 583 |
+
- contemplating
|
| 584 |
+
# Time-deferral with numbers — anti-bleed for COUNTER_SIZE
|
| 585 |
+
- give me 2 minutes
|
| 586 |
+
- back in 10
|
| 587 |
+
- wait 5
|
| 588 |
+
- in 3 mins
|
| 589 |
+
- 10 mins
|
| 590 |
+
- back in 15
|
| 591 |
+
- need a few mins
|
| 592 |
+
- need 2 minutes
|
| 593 |
+
# End-of-X time deferrals
|
| 594 |
+
- tomorrow
|
| 595 |
+
- after open
|
| 596 |
+
- after close
|
| 597 |
+
- by eod
|
| 598 |
+
- end of week
|
| 599 |
+
- after the meeting
|
| 600 |
+
- after this call
|
| 601 |
+
- first thing tomorrow
|
| 602 |
+
|
| 603 |
+
COUNTER_SIZE:
|
| 604 |
+
core:
|
| 605 |
+
- size 10
|
| 606 |
+
- size 5
|
| 607 |
+
- size 20
|
| 608 |
+
- 10 shares
|
| 609 |
+
- 15 sh
|
| 610 |
+
- resize to 10
|
| 611 |
+
- resize 12
|
| 612 |
+
- make it 15
|
| 613 |
+
- do 8
|
| 614 |
+
- do 5 shares
|
| 615 |
+
- shares 10
|
| 616 |
+
- cut to 10
|
| 617 |
+
- reduce to 5
|
| 618 |
+
- shrink to 8
|
| 619 |
+
- half size
|
| 620 |
+
- half
|
| 621 |
+
- double
|
| 622 |
+
- bump to 25
|
| 623 |
+
- try 12 shares
|
| 624 |
+
- 7 sh instead
|
| 625 |
+
- size at 10
|
| 626 |
+
- resize 15 sh
|
| 627 |
+
# "min N" / "max N" / "floor N" / "ceiling N" — anti-bleed from time
|
| 628 |
+
- min 3
|
| 629 |
+
- max 5
|
| 630 |
+
- min 8 shares
|
| 631 |
+
- max 12 sh
|
| 632 |
+
- minimum 5
|
| 633 |
+
- maximum 8
|
| 634 |
+
- floor 5
|
| 635 |
+
- ceiling 10
|
| 636 |
+
- cap at 8
|
| 637 |
+
- cap 10
|
| 638 |
+
# Dump / trim / load / exit / sell / buy verbs (with numbers or "half")
|
| 639 |
+
- dump half
|
| 640 |
+
- dump 50
|
| 641 |
+
- dump 65
|
| 642 |
+
- dump some
|
| 643 |
+
- dump a third
|
| 644 |
+
- trim half
|
| 645 |
+
- trim 50
|
| 646 |
+
- trim some
|
| 647 |
+
- trim 25
|
| 648 |
+
- load half
|
| 649 |
+
- load 50
|
| 650 |
+
- load up some
|
| 651 |
+
- load a third
|
| 652 |
+
- exit half
|
| 653 |
+
- exit 50
|
| 654 |
+
- sell half
|
| 655 |
+
- sell 50
|
| 656 |
+
- buy half
|
| 657 |
+
- buy 50
|
| 658 |
+
# Scale verbs (directional resize)
|
| 659 |
+
- scale in
|
| 660 |
+
- scale out
|
| 661 |
+
- scale in halfway
|
| 662 |
+
- scale out halfway
|
| 663 |
+
- scale in to 10
|
| 664 |
+
- scale out to 5
|
| 665 |
+
# Quantifier-only ("fewer" / numeric-less) — anti-bleed for "more"/"less" eval rows
|
| 666 |
+
- fewer
|
| 667 |
+
- fewer shares
|
| 668 |
+
- quite a bit less
|
| 669 |
+
- way more
|
| 670 |
+
- bit more
|
| 671 |
+
- try fewer
|
| 672 |
+
adversarial:
|
| 673 |
+
- approve at 10 shares
|
| 674 |
+
- yes but size 10
|
| 675 |
+
- ok do 5 instead
|
| 676 |
+
- go with 12
|
| 677 |
+
# Multi-verb COUNTER_SIZE — verb + size in one phrase
|
| 678 |
+
- dump half position
|
| 679 |
+
- dump half the position
|
| 680 |
+
- trim half position
|
| 681 |
+
- exit half the position
|
| 682 |
+
- sell half the position
|
| 683 |
+
- cut the position to 5
|
| 684 |
+
- reduce position to 10
|
| 685 |
+
# "approve at N" / "yes N" / "ok N" — approve + number compound
|
| 686 |
+
- approve at 5
|
| 687 |
+
- approve at 8
|
| 688 |
+
- approve at 12
|
| 689 |
+
- do it at 7
|
| 690 |
+
- yes at 8
|
| 691 |
+
- yes 5
|
| 692 |
+
- yes 7
|
| 693 |
+
- yes 12
|
| 694 |
+
- ok 5
|
| 695 |
+
- ok 7
|
| 696 |
+
- ok 12
|
| 697 |
+
- alright 10
|
| 698 |
+
- alright 8
|
| 699 |
+
- yep 9
|
| 700 |
+
- yep 11
|
| 701 |
+
- yeah 6
|
| 702 |
+
# ─── Balancing additions (msg 908) — broader size verbiage ─────
|
| 703 |
+
# "make it N"
|
| 704 |
+
- make it 4
|
| 705 |
+
- make it 6
|
| 706 |
+
- make it 7
|
| 707 |
+
- make it 8
|
| 708 |
+
- make it 9
|
| 709 |
+
- make it 11
|
| 710 |
+
- make it 13
|
| 711 |
+
- make it 16
|
| 712 |
+
- make it 18
|
| 713 |
+
# "do N" / "do N shares"
|
| 714 |
+
- do 4
|
| 715 |
+
- do 6
|
| 716 |
+
- do 7
|
| 717 |
+
- do 9
|
| 718 |
+
- do 11
|
| 719 |
+
- do 13
|
| 720 |
+
- do 14
|
| 721 |
+
- do 4 shares
|
| 722 |
+
- do 6 shares
|
| 723 |
+
- do 7 shares
|
| 724 |
+
- do 9 shares
|
| 725 |
+
- do 11 shares
|
| 726 |
+
# "use N" / "try N"
|
| 727 |
+
- use 8
|
| 728 |
+
- use 10
|
| 729 |
+
- use 12
|
| 730 |
+
- use 6 shares
|
| 731 |
+
- try 6
|
| 732 |
+
- try 11
|
| 733 |
+
- try 12
|
| 734 |
+
- try 14
|
| 735 |
+
- try 8 shares
|
| 736 |
+
# "go N" / "lets do N"
|
| 737 |
+
- go 4
|
| 738 |
+
- go 5
|
| 739 |
+
- go 6
|
| 740 |
+
- go 8
|
| 741 |
+
- go 9
|
| 742 |
+
- go 11
|
| 743 |
+
- lets try 6
|
| 744 |
+
- lets do 9
|
| 745 |
+
- lets use 10
|
| 746 |
+
- lets do 5 sh
|
| 747 |
+
- lets do 6 sh
|
| 748 |
+
- lets do 8 sh
|
| 749 |
+
# More "size N" / "N sh" / "N shares"
|
| 750 |
+
- size 4
|
| 751 |
+
- size 6
|
| 752 |
+
- size 7
|
| 753 |
+
- size 9
|
| 754 |
+
- size 11
|
| 755 |
+
- size 13
|
| 756 |
+
- 4 sh
|
| 757 |
+
- 6 sh
|
| 758 |
+
- 9 sh
|
| 759 |
+
- 11 sh
|
| 760 |
+
- 13 sh
|
| 761 |
+
- 4 shares
|
| 762 |
+
- 6 shares
|
| 763 |
+
- 9 shares
|
| 764 |
+
- 11 shares
|
| 765 |
+
- 13 shares
|
| 766 |
+
# "actually N" / "N not M"
|
| 767 |
+
- actually 7
|
| 768 |
+
- actually 6
|
| 769 |
+
- actually 12
|
| 770 |
+
- 5 not 10
|
| 771 |
+
- 10 not 15
|
| 772 |
+
- 12 not 20
|
| 773 |
+
- 8 not 15
|
| 774 |
+
# "swap/switch/redo at N"
|
| 775 |
+
- swap to 6
|
| 776 |
+
- swap to 8
|
| 777 |
+
- switch to 6
|
| 778 |
+
- switch to 8
|
| 779 |
+
- redo at 8
|
| 780 |
+
- redo at 5
|
| 781 |
+
# Half / fractional forms (not in eval)
|
| 782 |
+
- trim to half
|
| 783 |
+
- cut to half
|
| 784 |
+
- reduce to half
|
| 785 |
+
- do half size
|
| 786 |
+
- do half shares
|
| 787 |
+
- buy half
|
| 788 |
+
- sell half
|
| 789 |
+
- buy 50%
|
| 790 |
+
- sell 50%
|
| 791 |
+
|
| 792 |
+
COUNTER_PRICE:
|
| 793 |
+
# User wants to execute but at a different limit price (not share count).
|
| 794 |
+
# Positive signals: "$", "limit", "price", "@$N", verb + "at $N".
|
| 795 |
+
# Disambiguator: presence of $ / keyword pushes to COUNTER_PRICE;
|
| 796 |
+
# plain number (no $/limit/price) stays COUNTER_SIZE.
|
| 797 |
+
core:
|
| 798 |
+
# Dollar-sign explicit
|
| 799 |
+
- at $49
|
| 800 |
+
- at $50
|
| 801 |
+
- at $48
|
| 802 |
+
- at $42
|
| 803 |
+
- "@$49"
|
| 804 |
+
- "@$50"
|
| 805 |
+
- "@$42"
|
| 806 |
+
- $49 limit
|
| 807 |
+
- $50 limit
|
| 808 |
+
- $42 limit
|
| 809 |
+
- limit $49
|
| 810 |
+
- limit $50
|
| 811 |
+
- price $49
|
| 812 |
+
- price $50
|
| 813 |
+
# "limit X" / "price X" — keyword form (no $ needed)
|
| 814 |
+
- limit 49
|
| 815 |
+
- limit 50
|
| 816 |
+
- limit 42
|
| 817 |
+
- limit at 49
|
| 818 |
+
- limit at 50
|
| 819 |
+
- price 49
|
| 820 |
+
- price 50
|
| 821 |
+
- set limit at 49
|
| 822 |
+
- set the limit at 50
|
| 823 |
+
- raise limit to 50
|
| 824 |
+
- lower limit to 48
|
| 825 |
+
- change limit to 50
|
| 826 |
+
- move limit to 49
|
| 827 |
+
- new limit 50
|
| 828 |
+
- bump limit to 50
|
| 829 |
+
- drop limit to 48
|
| 830 |
+
# Verb + at + $N (dollar-explicit verb forms)
|
| 831 |
+
- trim at $49
|
| 832 |
+
- trim at $50
|
| 833 |
+
- exit at $50
|
| 834 |
+
- exit at $49
|
| 835 |
+
- sell at $49
|
| 836 |
+
- sell at $50
|
| 837 |
+
- buy at $42
|
| 838 |
+
- buy at $43
|
| 839 |
+
- close at $50
|
| 840 |
+
- dump at $49
|
| 841 |
+
- load at $42
|
| 842 |
+
- trim at $48.50
|
| 843 |
+
- exit at $49.50
|
| 844 |
+
- sell at $48.75
|
| 845 |
+
# "make it $X" / "do it at $X" / "execute at $X"
|
| 846 |
+
- make it $49
|
| 847 |
+
- make it $50
|
| 848 |
+
- do it at $49
|
| 849 |
+
- do it at $50
|
| 850 |
+
- run it at $50
|
| 851 |
+
- execute at $49
|
| 852 |
+
- execute at $50
|
| 853 |
+
- fire at $49
|
| 854 |
+
adversarial:
|
| 855 |
+
# Verb + at + plain number — relies on verb context to imply price
|
| 856 |
+
# over size. Eval has "trim at 49" — these should generalize.
|
| 857 |
+
- trim at 49
|
| 858 |
+
- trim at 50
|
| 859 |
+
- exit at 50
|
| 860 |
+
- sell at 49
|
| 861 |
+
- close at 50
|
| 862 |
+
- dump at 49
|
| 863 |
+
- load at 42
|
| 864 |
+
# COUNTER_PRICE with typos (mirror trial 5 typo)
|
| 865 |
+
- trim at 49 insteadd of
|
| 866 |
+
- exit at 50 insteadd
|
| 867 |
+
- sell at 49 instaed
|
| 868 |
+
- change price to 50
|
| 869 |
+
- new price 49
|
| 870 |
+
# "raise to $50" / "drop to $48" patterns
|
| 871 |
+
- raise to $50
|
| 872 |
+
- drop to $48
|
| 873 |
+
- raise to 50
|
| 874 |
+
- drop to 48
|
| 875 |
+
# ─── Balancing additions (msg 908) — broader $ explicit ─────
|
| 876 |
+
- at $51
|
| 877 |
+
- at $52
|
| 878 |
+
- at $46
|
| 879 |
+
- at $45
|
| 880 |
+
- at $44
|
| 881 |
+
- at $43
|
| 882 |
+
- at $48
|
| 883 |
+
- at $47
|
| 884 |
+
- $51 limit
|
| 885 |
+
- $52 limit
|
| 886 |
+
- $47 limit
|
| 887 |
+
- $46 limit
|
| 888 |
+
- $45 limit
|
| 889 |
+
- $44 limit
|
| 890 |
+
- $48 limit
|
| 891 |
+
- $48.50 limit
|
| 892 |
+
- $49.50 limit
|
| 893 |
+
- $50.25 limit
|
| 894 |
+
- $50.75 limit
|
| 895 |
+
- $42.50 limit
|
| 896 |
+
- $43.25 limit
|
| 897 |
+
# More limit/price keyword variations
|
| 898 |
+
- limit 51
|
| 899 |
+
- limit 52
|
| 900 |
+
- limit 53
|
| 901 |
+
- limit 47
|
| 902 |
+
- limit 46
|
| 903 |
+
- limit 45
|
| 904 |
+
- limit 48.5
|
| 905 |
+
- limit 49.25
|
| 906 |
+
- limit 50.5
|
| 907 |
+
- price 51
|
| 908 |
+
- price 52
|
| 909 |
+
- price 48
|
| 910 |
+
- price 47
|
| 911 |
+
- price 46
|
| 912 |
+
# Verb + at + $ (more prices, more verbs)
|
| 913 |
+
- trim at $50
|
| 914 |
+
- trim at $51
|
| 915 |
+
- trim at $52
|
| 916 |
+
- trim at $47
|
| 917 |
+
- trim at $46
|
| 918 |
+
- exit at $51
|
| 919 |
+
- exit at $52
|
| 920 |
+
- exit at $48
|
| 921 |
+
- exit at $47
|
| 922 |
+
- exit at $46
|
| 923 |
+
- sell at $50
|
| 924 |
+
- sell at $51
|
| 925 |
+
- sell at $47
|
| 926 |
+
- sell at $46
|
| 927 |
+
- sell at $45
|
| 928 |
+
- buy at $43
|
| 929 |
+
- buy at $44
|
| 930 |
+
- buy at $41
|
| 931 |
+
- buy at $40
|
| 932 |
+
- close at $51
|
| 933 |
+
- close at $48
|
| 934 |
+
- close at $47
|
| 935 |
+
- close at $46
|
| 936 |
+
- dump at $50
|
| 937 |
+
- dump at $51
|
| 938 |
+
- dump at $48
|
| 939 |
+
- dump at $46
|
| 940 |
+
- load at $43
|
| 941 |
+
- load at $41
|
| 942 |
+
- load at $40
|
| 943 |
+
- load at $44
|
| 944 |
+
# "the limit" / "the price" forms
|
| 945 |
+
- raise the limit
|
| 946 |
+
- lower the limit
|
| 947 |
+
- tighten the limit
|
| 948 |
+
- loosen the limit
|
| 949 |
+
- adjust the limit
|
| 950 |
+
- move the limit
|
| 951 |
+
- update the limit
|
| 952 |
+
- modify the limit
|
| 953 |
+
- shift the limit
|
| 954 |
+
- shift limit up
|
| 955 |
+
- shift limit down
|
| 956 |
+
- shift to $49
|
| 957 |
+
- shift to $50
|
| 958 |
+
- tighten to $48
|
| 959 |
+
- tighten to $49
|
| 960 |
+
- loosen to $50
|
| 961 |
+
- loosen to $51
|
| 962 |
+
# "rather" / "I want" / "give it" forms
|
| 963 |
+
- rather at $48
|
| 964 |
+
- rather at $50
|
| 965 |
+
- rather $49 limit
|
| 966 |
+
- I want $49
|
| 967 |
+
- I want it at $49
|
| 968 |
+
- I want at $48
|
| 969 |
+
- give it $49
|
| 970 |
+
- give it $50
|
| 971 |
+
# "let's do $X" / "go $X" — bridging APPROVE phrasing
|
| 972 |
+
- let's do $48
|
| 973 |
+
- let's do $50
|
| 974 |
+
- let's do $51
|
| 975 |
+
- go $49
|
| 976 |
+
- go $50
|
| 977 |
+
- go $48
|
| 978 |
+
# "$X instead" forms — bridging msg 898's "instead of"
|
| 979 |
+
- $48 instead
|
| 980 |
+
- $50 instead
|
| 981 |
+
- $51 instead
|
| 982 |
+
- $48 instead of 50
|
| 983 |
+
- $50 instead of 48
|
| 984 |
+
|
| 985 |
+
CROSSCLASS_NEGATIVES:
|
| 986 |
+
# These look like a clean label but are NOT safe to auto-commit.
|
| 987 |
+
# The model is trained to classify all of these as UNCLEAR.
|
| 988 |
+
# Falls through to Tier 2 confirmation DM.
|
| 989 |
+
multi_intent:
|
| 990 |
+
# NB: ticker-bearing multi-intent seeds are NOT here — they are
|
| 991 |
+
# template-generated in scripts/classifier/gen_synth_data.py via
|
| 992 |
+
# TICKER_POOL × MULTI_INTENT_TEMPLATES for broader variety than
|
| 993 |
+
# hand-writing could provide.
|
| 994 |
+
- yes but cut size in half
|
| 995 |
+
- approve and pause the others
|
| 996 |
+
- approve but check the chart first
|
| 997 |
+
- ok but smaller
|
| 998 |
+
- decline this and the next one
|
| 999 |
+
followup_questions:
|
| 1000 |
+
- whats the entry?
|
| 1001 |
+
- whats the stop?
|
| 1002 |
+
- why this size?
|
| 1003 |
+
- what triggered this?
|
| 1004 |
+
- which level?
|
| 1005 |
+
- is the chart still clean?
|
| 1006 |
+
- did the news change?
|
| 1007 |
+
partial:
|
| 1008 |
+
# Pure no-position fillers (NOT active opinion-forming — those are HOLD now)
|
| 1009 |
+
- hmm
|
| 1010 |
+
- not sure
|
| 1011 |
+
offtopic:
|
| 1012 |
+
- nice
|
| 1013 |
+
- cool
|
| 1014 |
+
- thanks
|
| 1015 |
+
- lol
|
| 1016 |
+
- "?"
|
| 1017 |
+
- "??"
|
| 1018 |
+
- whats new
|
| 1019 |
+
- status
|
| 1020 |
+
- dashboard link
|
| 1021 |
+
- any other ideas
|
| 1022 |
+
sarcasm_negation:
|
| 1023 |
+
- sure approve another loser
|
| 1024 |
+
- oh great another one
|
| 1025 |
+
- yeah right
|
| 1026 |
+
- of course not
|
| 1027 |
+
- yeah no
|
| 1028 |
+
- no yeah
|
| 1029 |
+
typo_traps:
|
| 1030 |
+
- aprove
|
| 1031 |
+
- dcline
|
| 1032 |
+
- hodl
|
| 1033 |
+
- holdd
|
| 1034 |
+
- aprove it
|
| 1035 |
+
ambiguous_size:
|
| 1036 |
+
- smaller
|
| 1037 |
+
- bigger
|
| 1038 |
+
- less
|
| 1039 |
+
- more
|
| 1040 |
+
- tighter
|
| 1041 |
+
- looser
|
| 1042 |
+
|
| 1043 |
+
UNCLEAR_core:
|
| 1044 |
+
# Genuinely UNCLEAR — no extractable position. User is either asking
|
| 1045 |
+
# for info, off-topic, evaluative without deciding, or expressing pure
|
| 1046 |
+
# filler/noise. State-of-attention phrases ("looking", "brb") moved to
|
| 1047 |
+
# HOLD per the active-deferral framing; THIS pile is the "I'm not even
|
| 1048 |
+
# taking a position to defer" set.
|
| 1049 |
+
core:
|
| 1050 |
+
# Filler / non-committal noise
|
| 1051 |
+
- hm
|
| 1052 |
+
- uhh
|
| 1053 |
+
- umm
|
| 1054 |
+
- hmm ok
|
| 1055 |
+
- ehh
|
| 1056 |
+
- mehhh
|
| 1057 |
+
- dunno
|
| 1058 |
+
- shrug
|
| 1059 |
+
# Requests for info / off-topic
|
| 1060 |
+
- send me the link
|
| 1061 |
+
- what was the price again
|
| 1062 |
+
- did this fire already
|
| 1063 |
+
- is this the same as yesterday
|
| 1064 |
+
- whats the trigger
|
| 1065 |
+
- hows the portfolio
|
| 1066 |
+
- whats new
|
| 1067 |
+
- show me the levels
|
| 1068 |
+
- share the chart
|
| 1069 |
+
- give me the brief
|
| 1070 |
+
- show options chain
|
| 1071 |
+
- pull options
|
| 1072 |
+
- what about iv
|
| 1073 |
+
# Evaluative-not-committal — describing the trade without deciding
|
| 1074 |
+
- nice chart
|
| 1075 |
+
- clean setup
|
| 1076 |
+
- decent risk reward
|
| 1077 |
+
- solid thesis
|
| 1078 |
+
- well structured
|
| 1079 |
+
- thoughtful
|
| 1080 |
+
- good analysis
|
| 1081 |
+
- neat trade
|
| 1082 |
+
- fair setup
|
| 1083 |
+
- reasonable
|
| 1084 |
+
- could work
|
| 1085 |
+
- i see it
|
| 1086 |
+
- i see
|
| 1087 |
+
- intriguing
|
| 1088 |
+
# Multi-intent / self-contradiction — can't safely commit
|
| 1089 |
+
- approve but also wait
|
| 1090 |
+
- yes and then no
|
| 1091 |
+
- do it but not yet
|
| 1092 |
+
- execute then trim
|
| 1093 |
+
- skip and find another
|
| 1094 |
+
- yes wait actually no
|
| 1095 |
+
- approve another not this
|
| 1096 |
+
# Single-letter / two-letter ambiguity — UNCLEAR
|
| 1097 |
+
- j
|
| 1098 |
+
- m
|
| 1099 |
+
- z
|
| 1100 |
+
- p
|
| 1101 |
+
- ohh
|
| 1102 |
+
- aah
|
| 1103 |
+
- ehhh
|
| 1104 |
+
- ok thx
|
| 1105 |
+
# "looks X" evaluative — explicitly not a decision
|
| 1106 |
+
- looks fine
|
| 1107 |
+
- looks alright
|
| 1108 |
+
- looks ok
|
| 1109 |
+
- appears good
|
| 1110 |
+
- seems ok
|
| 1111 |
+
- feels right
|
| 1112 |
+
- this feels solid
|
| 1113 |
+
- looks legit
|
| 1114 |
+
- seems fine
|
| 1115 |
+
- appears solid
|
| 1116 |
+
# "idk" + uncertainty
|
| 1117 |
+
- idk
|
| 1118 |
+
- no idea
|
| 1119 |
+
- im not certain
|
| 1120 |
+
- im torn
|
| 1121 |
+
- couldnt tell you
|
| 1122 |
+
- im undecided
|
| 1123 |
+
- cant decide
|
| 1124 |
+
- havent decided
|
| 1125 |
+
- im split on this
|
| 1126 |
+
# "your call" / "you decide" — handing the decision back to the agent
|
| 1127 |
+
- your call
|
| 1128 |
+
- you call
|
| 1129 |
+
- u decide
|
| 1130 |
+
- you decide
|
| 1131 |
+
- up to you
|
| 1132 |
+
- up to the engine
|
| 1133 |
+
- whatever you think
|
| 1134 |
+
- whatever you decide
|
| 1135 |
+
- agent decides
|
| 1136 |
+
- engine call
|
| 1137 |
+
- your choice
|
| 1138 |
+
- whatever works
|
| 1139 |
+
# "idk + you" compound — CRITICAL safety class.
|
| 1140 |
+
# These all signal "I don't know, the engine should pick" → UNCLEAR.
|
| 1141 |
+
# Run #7 had "idk you call it" → APPROVE @ 0.91 (high-conf wrong commit).
|
| 1142 |
+
- idk you make the call
|
| 1143 |
+
- idk your decision
|
| 1144 |
+
- idk your pick
|
| 1145 |
+
- no clue you decide
|
| 1146 |
+
- no idea you decide
|
| 1147 |
+
- you make the call
|
| 1148 |
+
- you make the choice
|
| 1149 |
+
- make the call yourself
|
| 1150 |
+
- engine pick
|
| 1151 |
+
- bot pick
|
| 1152 |
+
- bot decides
|
| 1153 |
+
- dunno you pick
|
| 1154 |
+
- dunno you decide
|
| 1155 |
+
- idk pick for me
|
| 1156 |
+
- cant decide you pick
|
| 1157 |
+
- whatever the engine wants
|
| 1158 |
+
# Multi-intent / self-contradiction stronger — "no wait yes" @ 0.92 wrong
|
| 1159 |
+
- no actually yes
|
| 1160 |
+
- yes wait no
|
| 1161 |
+
- no then yes
|
| 1162 |
+
- approve actually no
|
| 1163 |
+
- decline actually yes
|
| 1164 |
+
- hold no wait
|
| 1165 |
+
- yes hold no
|
| 1166 |
+
- approve no decline
|
| 1167 |
+
- back and forth on this
|
| 1168 |
+
- flip flopping
|
| 1169 |
+
- mind changed back
|
| 1170 |
+
# Hand-back to bot — alternate "idk you call it" forms
|
| 1171 |
+
- dont know up to you
|
| 1172 |
+
- couldnt decide u pick
|
| 1173 |
+
- you tell me
|
| 1174 |
+
- hard to say you choose
|
| 1175 |
+
- im split agent decide
|
| 1176 |
+
- im torn engine pick
|
| 1177 |
+
# ─── Balancing additions (msg 908) — UNCLEAR variety ─────
|
| 1178 |
+
- uhmm
|
| 1179 |
+
- hmmmm
|
| 1180 |
+
- eh
|
| 1181 |
+
- eh idk
|
| 1182 |
+
- no comment
|
| 1183 |
+
- explain more
|
| 1184 |
+
- tell me more
|
| 1185 |
+
- what does that mean
|
| 1186 |
+
- say again
|
| 1187 |
+
- huh
|
| 1188 |
+
- huh what
|
| 1189 |
+
- wait what
|
| 1190 |
+
- wait i dont follow
|
| 1191 |
+
- lost me
|
| 1192 |
+
- i dont follow
|
| 1193 |
+
- im not following
|
| 1194 |
+
- which is this one
|
training.jsonl
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|