Prakash Chaudhary connectwithprakash

Skill Eval Benchmark Results

Date: 2026-04-02 PR: https://github.com/fetch-rewards/labs-skills/pull/16 Skills: summarizing-specs, gathering-coding-context

Skill	Pass Rate	Baseline	Lift	Token Delta

A self-contained benchmark comparing two LLM tool-calling patterns across queries of varying complexity (1 tool through 8+ tools).

Tool-as-Code: LLM generates a complete Python function in one shot, executes locally. Tool results never re-enter the LLM context.
ReAct: Standard agent loop -- LLM calls tools one at a time, reads each result, decides the next step.

Both use GPT-4.1 with the same 20 tools. The benchmark measures tokens, latency, and LLM round-trips.

	def check_constraints(constraints, image_dim, window_axes):
	# checks the movement of axes possible or not
	move_left, move_right, move_up, move_down = constraints
	x, y = image_dim
	x1, x2, y1, y2 = window_axes
	if x1<=0: move_left = False
	if x2>=x: move_right = False
	if y1<=0: move_up = False
	if y2>=y: move_down = False

	detector = MTCNN() # face detection model
	image = plt.imread(input_file_path)

	results = detector.detect_faces(image)
	confidences = [result['confidence'] for result in results]
	index = np.argmax(confidences)
	x1, y1, ww, hh = results[index]['box']
	h, k = x1+ww//2, y1+hh//2

	image = crop(image, (h, k), aspect_ratio)

	image = plt.imread(file_path)
	m, n = image.shape[0:-1]
	image = crop(image, center=(m//2+290, n//2), aspect_ratio=(1.4/2.0))
	plt.imsave('output.jpg', image)

	"""
	Minimal character-level Vanilla RNN model. Written by Andrej Karpathy (@karpathy)
	BSD License
	"""
	import numpy as np

	# data I/O
	data = open('input.txt', 'r').read() # should be simple plain text file
	chars = list(set(data))
	data_size, vocab_size = len(data), len(chars)

	def get_filter_output(input_img_path, filter):
	img = plt.imread(input_img_path)
	filter_size = filter.shape[0]
	pad_width = (filter_size-1)//2
	if len(img.shape)==3:
	filter_3D = np.stack((filter, filter, filter), axis=2)
	temp = np.pad(img, ((pad_width, pad_width), (pad_width, pad_width), (0, 0)), mode='constant', constant_values=0)
	temp = np.array([[np.sum(np.multiply(temp[i:i+filter_size, j:j+filter_size, :], filter_3D)) for j in range(temp.shape[1]-filter_size+1)] for i in range(temp.shape[0]-filter_size+1)])
	else:
	temp = np.pad(img, ((pad_width, pad_width), (pad_width, pad_width)), mode='constant', constant_values=0)