question string | metadata string | task_source string | formatted_prompt list | response list | response__extractions list | response__extractions_metadata list | eval_correct list | strategy_compliance int64 | compliance_reasoning string | compliance_elements_followed list | compliance_elements_missed list |
|---|---|---|---|---|---|---|---|---|---|---|---|
Using ALL of the numbers 18, 3, 14, 87, 95, 70, 12, 76, create an expression that equals 196. You must use each number exactly once. You may use +, -, *, /. | {"numbers": [18, 3, 14, 87, 95, 70, 12, 76], "target": 196, "operators_used": ["/", "+", "+", "+", "-", "-", "+"], "expression": "18 / 3 + 14 + 87 + 95 - 70 - 12 + 76"} | countdown | [
{
"content": "You are a math problem solver. Solve the given arithmetic problem step by step. Show your work and then provide your final answer inside <answer> tags.\n\nFor example: <answer>10 + 5 * 2 = 20</answer>\n\nThe answer should be a single arithmetic expression using the given numbers that equals the ta... | [
"We are given the numbers: \n**18, 3, 14, 87, 95, 70, 12, 76** \nWe must use **each exactly once** and combine them with `+`, `-`, `*`, `/` to make **196**.\n\n---\n\n### Step 1: Look at the total sum of all numbers\nLet’s first compute the sum of all numbers to get a sense of scale:\n\n18 + 3 = 21 \n21 + 14 = 3... | [
[
"95 + 87 + 14 + (76 + 12 - 70 - 18) * 3 = 196"
]
] | [
[
{
"end_index": 2691,
"end_phrase": "</answer>",
"start_index": 2630,
"start_phrase": "<answer>"
}
]
] | [
[
false
]
] | 2 | The model begins with an inventory of numbers and the target, but quickly deviates from the systematic tree search strategy. It does not explicitly branch through pairs of numbers or operations, nor does it track the available number set after each operation. The reasoning is more heuristic and exploratory rather than ... | [
"Lists all available numbers and target at the start (INVENTORY step)"
] | [
"Picks pairs of numbers and tries operations systematically (BRANCH step)",
"Tracks the available number set after each operation",
"Uses explicit tree structure (branches, levels, indentation, or labeling)",
"Explicitly marks dead ends (writes DEAD END or BACKTRACK or equivalent)",
"Returns to previous bra... |
Using ALL of the numbers 77, 42, 63, 3, 15, 45, 40, 31, create an expression that equals 176. You must use each number exactly once. You may use +, -, *, /. | {"numbers": [77, 42, 63, 3, 15, 45, 40, 31], "target": 176, "operators_used": ["+", "*", "+", "+", "/", "+", "-"], "expression": "77 + 42 * 63 + 3 + 15 / 45 + 40 - 31"} | countdown | [
{
"content": "You are a math problem solver. Solve the given arithmetic problem step by step. Show your work and then provide your final answer inside <answer> tags.\n\nFor example: <answer>10 + 5 * 2 = 20</answer>\n\nThe answer should be a single arithmetic expression using the given numbers that equals the ta... | [
"We are given the numbers: \n**77, 42, 63, 3, 15, 45, 40, 31** \nWe must use **each exactly once** and combine them with `+`, `-`, `*`, `/` to get **176**.\n\n---\n\n### Step 1: Look at the total sum of all numbers\nLet’s first add them all up to see how far we are from 176.\n\n77 + 42 = 119 \n119 + 63 = 182 \n... | [
[
"77 + 40 + 45 + 42 / 3 = 176",
"77 + 42 + 63 - (40 - 31 - 3) + (45 / 15 - 3) = 176"
]
] | [
[
{
"end_index": 20916,
"end_phrase": "</answer>",
"start_index": 20872,
"start_phrase": "<answer>"
},
{
"end_index": 22133,
"end_phrase": "</answer>",
"start_index": 22066,
"start_phrase": "<answer>"
}
]
] | [
[
false,
false
]
] | 2 | The model acknowledges the need to use all numbers and the target but quickly reverts to a more heuristic approach rather than systematically applying the tree search strategy. While it attempts to explore combinations, it does not consistently follow the structured steps of the prescribed strategy, leading to a lack o... | [
"Lists all available numbers and target at the start",
"Picks pairs of numbers and tries operations systematically"
] | [
"Tracks the available number set after each operation",
"Uses explicit tree structure",
"Explicitly marks dead ends",
"Returns to previous branch point after dead ends",
"Avoids heuristic reasoning",
"Shows solution path tracing back through successful branches"
] |
Using ALL of the numbers 53, 60, 1, 7, 87, 84, 83, 13, create an expression that equals 698. You must use each number exactly once. You may use +, -, *, /. | {"numbers": [53, 60, 1, 7, 87, 84, 83, 13], "target": 698, "operators_used": ["+", "/", "*", "+", "-", "-", "-"], "expression": "53 + 60 / 1 * 7 + 87 - 84 - 83 - 13"} | countdown | [
{
"content": "You are a math problem solver. Solve the given arithmetic problem step by step. Show your work and then provide your final answer inside <answer> tags.\n\nFor example: <answer>10 + 5 * 2 = 20</answer>\n\nThe answer should be a single arithmetic expression using the given numbers that equals the ta... | [
"We are given the numbers: \n**53, 60, 1, 7, 87, 84, 83, 13** \nWe must use **each exactly once**, and combine them with **+, -, *, /** to get **698**.\n\n---\n\n### Step 1: Estimate the target and look for large numbers\n\nTarget: **698**\n\nLarge numbers in the list: \n- 87, 84, 83, 60, 53, 13, 7, 1\n\nSum of ... | [
[
"87 * (7 + 1) + (84 - 83) + 1 = 698",
"83 * (7 + 1) + 87 - 53 = 698",
"(87 + 13) * 7 - (84 - 83) - 1 = 698",
"(87 * 8) + (84 - 83) + 1 = 698"
]
] | [
[
{
"end_index": 29420,
"end_phrase": "</answer>",
"start_index": 29369,
"start_phrase": "<answer>"
},
{
"end_index": 31015,
"end_phrase": "</answer>",
"start_index": 30970,
"start_phrase": "<answer>"
},
{
"end_index": 32913,
"end_phras... | [
[
false,
false,
false,
false
]
] | 1 | The model completely disregards the prescribed strategy and relies on ad-hoc reasoning throughout its response. It does not systematically list available numbers, attempt operations in a structured manner, or follow the tree search format as instructed. | [] | [
"Lists all available numbers and target at the start (INVENTORY step)",
"Picks pairs of numbers and tries operations systematically (BRANCH step)",
"Tracks the available number set after each operation (shows remaining numbers)",
"Uses explicit tree structure (branches, levels, indentation, or labeling)",
"... |
t1-strategy-countdown-baseline-together_ai-qwen-qwen3-next-80b-a3b-inst-39a439f1
Strategy compliance evaluation on countdown arithmetic — baseline variant.
Model received no strategy instruction (baseline control). Judge still scores against tree search rubric. Compliance is scored by an LLM judge (1-5 Likert).
Results
| Metric | Value |
|---|---|
| pass@1 | 0.0000 |
| Strategy compliance (mean) | 1.67 |
| Strategy compliance (min) | 1 |
| Strategy compliance (max) | 2 |
| Total problems | 3 |
Compliance Distribution
| Score | Count | Meaning |
|---|---|---|
| 1 | 1 | Non-compliant |
| 2 | 2 | Minimal |
| 3 | 0 | Partial |
| 4 | 0 | Mostly compliant |
| 5 | 0 | Fully compliant |
Details
| Parameter | Value |
|---|---|
| Model | together_ai/Qwen/Qwen3-Next-80B-A3B-Instruct |
| Thinking | False |
| Temperature | 0.7 |
| Max tokens | 16384 |
| Judge model | openai/gpt-4o-mini |
| Num examples | 3 |
| Num args | 8 |
| Variant | baseline |
Usage
from datasets import load_dataset
ds = load_dataset("reasoning-degeneration-dev/t1-strategy-countdown-baseline-together_ai-qwen-qwen3-next-80b-a3b-inst-39a439f1", split="train")
scores = ds["strategy_compliance"]
print(f"Mean compliance: {sum(scores) / len(scores):.2f}")
Tracked in reasoning-degeneration-dev/PROJECT-MANIFEST
- Downloads last month
- 19