Dense Slides: Reinforcement Learning for Agents - Will Brown, ML Researcher at Morgan Stanley
Source Video
Reinforcement Learning for Agents - Will Brown, ML Researcher at Morgan Stanley
Method
This deck is slide-only. The existing captured video frame set supplies candidate frames, then local OpenCV rejects sponsor/title/speaker-only frames, crops visible slide surfaces, deduplicates, and saves the cropped slide images.
Cropped Visible Slides

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
content_slideconfidence0.98 - Text source: advanced OCR
rapidocr-live/bright-screen/opencv-adaptive. - OCR decision: ready — Dense two-column slide with small table/diagram text and multiple bullet groups.
Slide text:
LLM agents in 2025
"Reasoning" models = the future? Most LLMs are Chatbots o. Helpfut for interactive problem-solving. Great on Q&A benchmarks: Level 1 Stages of Artificial Intelligence' Chatbots, Al with conversational language
o. o. 01, 03, R1, Grok 3,.. Better at hard Q&A. Level 2. solving Reasoners, human-level problem
Best practices for "agents" Level3 Agents, systems that can take actions
Many chained LLM calls per task: Results are... OK? Prompt engineering + tooling Evals, ops, tool-use, human-in-the-loop: Level 5 Level 4: Innovators, Al that can aid In irventlon.. of an organization Organizations, Al that can do the work (OpenAl)

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
content_slideconfidence0.92 - Text source: advanced OCR
rapidocr-live/bright-screen/opencv-adaptive. - OCR decision: ready — Main slide is readable but mixed with stage camera footage and branding panels; OCR is better for the small text.
Slide text:
LLM agents in 2025
AIE Most LLMs are Chatbots "Reasoning" models = the future? Helpful for interactive problem-soiving Great on Q&A benchmarks Level 1 Intelligonco Stages of Artificial language Chatbots, Al with corrversational
。 01, 03, R1, Grok 3,... Better at hard Q&A Level 2 sohving Reasoners, human-level problem
Best practices for "agents" Level 3 Agents, systems that can take actions
o Prompt engineering + tooling Evals, ops, tool-use, human-in-the-loop Results are... OK? Many chained LLM calls per task Level5 Levei 4 Innovators, Al that can aid in invention Organizatlons, Al that can do the work of an organization (OpenAl)
AGENTENGINEERING
HTtPS:IIAI.ENgINEER

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
content_slideconfidence0.98 - Text source: advanced OCR
rapidocr-live/bright-screen/contrast. - OCR decision: ready — Text-heavy slide with multiple small diagrams and labels that are suitable for OCR.
Slide text:
How agentic are our agents? LLMCA Ovtput2 LLMCHF
Oetsot: Cate
Fai Cxit
Many agents are pipelines Low degree of autonomy Pipelines
Winning apps = tight feedback loops (mostly) Non-trivial engineering required LLM CalI Router LLMCAE1 LLMCall2
O IDEs (Cursor,Windsurf,Replit) LLMCall3
O Betterathard Q&A
Not many agents“do stuff"for 10+min
O Devin,Operator,Deep Research Actioo
Howdowemakemoreof these? Human LLMCall Environment
Justwait forbetter models?
Classical RL = optimize agent policy Stop Feedback Agents
(Anthropic)

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
content_slideconfidence0.98 - Text source: advanced OCR
rapidocr-live/bright-screen/opencv-adaptive. - OCR decision: ready — Dense slide with charts and small annotations; OCR will read the chart labels and bullet text better.
Slide text:
Lessons from DeepSeek
1 month ago: DeepSeek R1 paper + model
RL on strong base models works
RL with sparse rewards works
01-level performance = easier than expected.
Long CoT = emergent
Open-source Iis back (for now))
Several replication efforts
Can distill to' smaller models:
So... what next?:
: (DeepSeek, 25)

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
content_slideconfidence0.98 - Text source: advanced OCR
rapidocr-live/bright-screen/contrast. - OCR decision: ready — Mixed text-and-diagram slide with small labels and a process diagram that is better handled by OCR.
Slide text:
How reinforcement learning works
Key idea: explore and exploit Largo-scnle Roasoning-Oronted Reinforcement Learning
DeepSeek's GRPO DeepSoek-v3-Base DoepSeek-R1-Zero
Like PPO but simpler Less compute, same effect Group-RelativePolicy Optimization Tran Hodet checkgodnt undee.crainimg (Ma)a u Rule-basedverication acodeapfhon? ×
Main loop:
O Sample N responses per prompt
Computerewards foreach
O "Advantage estimates" (Jay Alammar,'25)
Update rule:better score =increase likelihood
Challenges: rewards? multi-turn?

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
content_slideconfidence0.98 - Text source: advanced OCR
rapidocr-live/bright-screen/contrast. - OCR decision: ready — Slide has a dense text inset and a chart; OCR is the right first-pass extraction method.
Slide text:
Lessons from OpenAl Deep Research
LLM agents can work for long multi-step tasks
Vibe check = impressive
Still struggles“out of distribution”
Big labswill keep making agents
Don't know full details, but..
How it works C05
Deepresearch was trained using end-to-end reinforcement
domains. Through that training, it learned to plan and execute a learning on hard browsing and reasoning tasks across a range of The more the model browses and thinks about what its browsing. MaxTool Calls
multi-step trajectory to find the data it needs, backtracking and the better it does, which is why giving it time to think is important.
reacting to real-time information where necessary. The model is (OpenAl.'25)

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
content_slideconfidence0.97 - Text source: agent_vision.
Slide text:
RL infrastructure
- RL training infra: exists, but mostly "RLHF"-style
- Want options besides "wait for big lab tuning APIs"
- How can we plan ahead?
- Unknowns:
- Cost to do RL for agentic tasks?
- How small can the models be?
- Generalization across tasks?
- How to design effective rewards?
- Opportunities:
- Open-source infra for DIY
- Services for agentic RL

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
demo_videoconfidence0.89 - Text source: advanced OCR
rapidocr-live/bright-screen/opencv-adaptive. - OCR decision: ready — Dense screenshots, code, and small chart labels.
Slide text:
grpo_demo.py
will brown @willccbb Jan 26 Sints scorte ay the oooonent in th ch GRPO self-correction on Lama-18:) Promoto te are elso giytn that the points scored by the ooponent in the changionshlg gase is hslf of UF's nurite? (2)2)0(1/2)(21②02) Sirpllfying tho ecuntlon, Co cet.
2 tr (24x
This laptification ss arong, Tho opponsnt cust have scored & core point!.
2 Z psjo?s aaru lin lutuoddo su1 ter4 1 160 150 130 This simplifits to: This tipllties to. So. they stcored 1/6 polnte esch In ttin first 2a get (24x2)0.(1/2)(21( x2/12 (24x2)12x 24x.12x03 12x0.2 9/1x
24 484 1167K tain/rewards'xcurxy t!an/compltion_length
will brown cwillccbb Jan 26 aduss 03 10 1
01

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
title_cardconfidence0.84 - Text source: advanced OCR
rapidocr-live/bright-screen/contrast. - OCR decision: ready — Collage slide with multiple embedded screenshots and small text labels.
Slide text:
grpo_demo.py
GRPOself-correctlon onLama-1B:) willbrown@willccbo·Jan26 Promote meare alsogiven that the points cored by the cgpoent In thecharoiohi gae Ishalf of ts nrite: (24x-2)=(1/2)(24x-2·2)
AIE pintswoeredtythtopooaeatlathtchang tr Sinplifying theequatioo,weget: (24x-2)+(2x-2)
gett This sinplification was wroog. The opponeat most have scored 2 sore oints.
This sinplifies to: This sinptities to: So,thoy score 1/6 points ach in their first 26gae. (24x-2)=(1/2)(24x-2·2) x1/6 x2/12 (24×2)12x 12x2 24·222
chInteipfirst24ges,
Q24 Et 484 sl67K trainrewaresx(urxy traincompletionjength
script: willbrown@willccboJan26
AGENT ENGINEERING
HTTPS://AI.ENGINEER

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
demo_videoconfidence0.92 - Text source: advanced OCR
rapidocr-live/bright-screen/opencv-adaptive. - OCR decision: ready — Dense multi-panel collage with screenshots, code, and small UI text.
Slide text:
grpo_demo.py (*st ctme 6 hours +0o willccbb / grpo demo.py
Hundreds of Billions of Dotsrs Later Teknlum (e/入) 0 F 0Teknlum 1n30 This Is the entkre code needed to reproduce Rs ko! 一 Irokeaso todry the fnst Google Cotb nourbook to tain a reuroning nrw RrrlorceTtrt Ltvning slgorithm trom DepSetk, GRpo. h lte Jo) uru Aiy+ uea nor tpo 6uuea-rod o pue La o uonpol houri you can trrhform I tery srul model, Oron O.s (Soo mlor Hto a tirty muth reusonirg mruchint. To Uhh dale, Iha b probubhy th? oda PiarTe-Csrl Langlais. tu Co'nor Pth Revisons 60 Str3 960 MFoks 313 s-smits/grpo- optuna
The notedook a based on a tcrigt by Wilm Brorn thst signifcan Cn Llaru 1B csptcitits for math. This i a typical rtirforcerent hs Iprurh bayta os tocLtrduo erriplt ton cie girn cCire! has to find the solution, resrron ovor n for sorme tine (tis is the sbiotle basic Torm ol inetnce time carmacte) tnd rtttding the gooa resronihg dirtr Trsnino I smat meth reasone wth F1 Qwen o Sb on GRPO Reinsorcemant Leunirg rRLi tree vning m poou buuo+res Lu-sse,dneg wao nof Lre1 ed ov Hog poet aha nha M-snarrg Reasoning-GRPO & RL
GSM8K Reward Functions
Srmng up iher mocdet
Ihet pto, Moy
2.3M cce3.763 80coTmm:·307ra90t1

- Recreated text/layout view: open HTML recreation
- AI slide classifier:
content_slideconfidence0.96 - Text source: agent_vision.
Slide text:
Rubric engineering
- An accessible way to steer RL algorithms
- Invites creativity + experimentation
- Anecdotally, seems to help
- The next "prompt engineering"?
- Next steps:
- Using LLMs to design rubrics?
- Auto-tuning rubrics with DSPy?
- Incorporating LLM judges for soft criteria?
- Avoiding reward hacking?
Classification audit: raw/sources/slide-ai-classification/dense/JIsgyk0Paic/audit.json